Perimeter intrusion alarm method and system

Abstract

The embodiment of the invention provides a perimeter intrusion alarm method and system. Wherein image features in the real-time environment interaction information are recognized through a convolutional neural network, so that a behavior pattern recognition result of the real-time environment interaction information is obtained; dynamically constructing a moving path of a potential intruder in the virtual three-dimensional geographic information system to generate a moving path estimation result; when an abnormal activity mode is detected, a multi-mode warning response mechanism is activated immediately, and an optimal warning response combination is obtained through a genetic algorithm; when a non-human biological intruder invades, a harmless repelling measure for the non-human biological intruder is generated through the eco-friendly interference device; and learning a group behavior mode by using the anonymized real-time environment interaction information, and continuously optimizing an optimal warning response combination and a harmless expelling measure. According to the technical scheme provided by the embodiment of the invention, the sensitivity and reliability of perimeter intrusion alarm can be improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of perimeter intrusion alarm technology, and more particularly to a perimeter intrusion alarm method and system. Background Art

[0002] In modern security protection systems, perimeter intrusion alarm methods are widely used in various critical infrastructures and sensitive areas, such as military bases, airports, nuclear power plants, industrial parks, etc. These places have extremely high requirements for security and reliability, and need to be able to monitor in real time and accurately identify potential intrusion behaviors.

[0003] Currently, perimeter intrusion alarm systems mainly rely on traditional sensor technology and rule-based alarm mechanisms. Common solutions include: installing fixed cameras and infrared sensors to monitor the perimeter, combined with simple motion detection algorithms to trigger alarms; using physical fences and vibration sensors to detect climbing or vandalism; using radar and acoustic wave detectors for long-distance target detection, combined with preset thresholds to determine whether to trigger an alarm.

[0004] Although existing solutions have met basic security needs to a certain extent, there are still many shortcomings: traditional sensor technology is difficult to cope with complex environmental changes and is prone to false alarms or missed alarms, affecting system reliability and user experience; most existing alarm systems adopt fixed response strategies, lack flexibility and adaptability, and are unable to dynamically adjust warning measures according to actual conditions, resulting in limited ability to respond to emergencies; many traditional expulsion methods may cause harm to non-human organisms and even damage the ecological environment, which does not meet the requirements of modern society for sustainable development. Summary of the invention

[0005] The embodiments of the present application provide a perimeter intrusion alarm method and system to solve the problems of low sensitivity and poor reliability of perimeter intrusion alarm in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a perimeter intrusion alarm method, comprising: Receive real-time environment interaction information from multiple distributed intelligent perception nodes, identify image features in the real-time environment interaction information through a convolutional neural network, and parse the time series behavior pattern in the real-time environment interaction information to obtain a behavior pattern recognition result of the real-time environment interaction information; According to the behavior pattern recognition result, dynamically construct the movement path of the potential intruder in the virtual three-dimensional geographic information system, and adjust the prediction method of the movement path in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; Based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm; Based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure is generated by an eco-friendly interference device to prevent the perimeter from being illegally crossed and to protect the ecological environment from being damaged; Based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern is learned by using the anonymized real-time environmental interaction information, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to improve the sensitivity of the perimeter intrusion alarm.

[0007] Optionally, based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, and the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm, including: Based on the movement path estimation result, the movement path estimation result is analyzed by using anomaly detection and time series prediction, and when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated to generate an immediate response measure for the abnormal activity pattern; According to the immediate response measures, combining the intelligence-driven decision-making framework with natural language processing technology, using graph neural networks and knowledge graphs to analyze the on-site conditions of the potential intruder, so as to obtain an assessment report of the potential intruder; Based on the evaluation report of the potential intruder, the corresponding warning response combination is searched using genetic algorithm and particle swarm optimization, and the reinforcement learning strategy is used to continuously adjust and optimize according to the real-time feedback information of the actual situation of the potential intruder to generate the best warning response combination.

[0008] Optionally, according to the immediate response measures, in combination with the intelligence-driven decision-making framework and natural language processing technology, the on-site conditions of the potential intruder are analyzed using a graph neural network and a knowledge graph to obtain an assessment report of the potential intruder, including: Integrate and analyze the multi-source intelligence information collected in the immediate response measures on the potential intruder using an intelligence-driven decision-making framework according to the immediate response measures to generate a site situation analysis report including perimeter intrusion site environmental information and real-time monitoring information; Based on the on-site situation analysis report, using natural language processing technology to analyze the unstructured multi-source intelligence information in the on-site situation analysis report, and extracting the key information of the potential intruder to generate structured description information of the key information; Based on the structured description information, using a graph neural network, updating the knowledge graph of the potential intruder in the structured description information to obtain an updated knowledge graph; According to the entity relationships in the updated knowledge graph, the on-site conditions including the identity, capabilities and behavior trajectory of the potential intruder are identified and parsed to obtain an assessment report of the potential intruder.

[0009] Optionally, according to the behavior pattern recognition result, the movement path of the potential intruder is dynamically constructed in the virtual three-dimensional geographic information system, and the prediction method of the movement path is adjusted in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result, including: According to the behavior pattern recognition result, the behavior pattern recognition result is analyzed by using a virtual three-dimensional geographic information system to obtain a movement path model of each potential intruder; Based on the movement path model, using Bayesian optimization, evaluating and optimizing the movement path model, generating a preliminary movement path prediction method, so as to improve the accuracy of the movement path prediction of the potential intruder; Based on the preliminary moving path prediction method, a scenario simulation analysis process is performed on the reinforcement learning algorithm, and the preliminary moving path prediction method is further adjusted to obtain the moving path prediction method; Based on the prediction method of the moving path, the correction information of the prediction method fed back by reinforcement learning is used to analyze and correct the uncertainty and error in the moving path model to generate a moving path estimation result.

[0010] Optionally, based on the preliminary movement path prediction method, a scenario simulation analysis process is performed on the reinforcement learning algorithm, and the preliminary movement path prediction method is further adjusted to obtain the movement path prediction method, including: Based on the preliminary movement path prediction method, the terrain conditions, weather conditions and emergency scenarios of the potential intruder are simulated to generate simulation scenarios including various movement paths; According to the simulation scenario, a reinforcement learning algorithm is applied to simulate the behavior pattern of the potential intruder, and the simulation result is compared and analyzed with the preset ideal behavior pattern of the potential intruder, the actual performance of the preliminary movement path prediction method is evaluated, and an analysis report of the preliminary movement path prediction method is obtained; Based on the analysis report, adjusting the reward mechanism and the penalty mechanism in the reinforcement learning algorithm to optimize the preliminary movement path prediction method and generate an improved movement path prediction method; Based on the improved mobile path prediction method, the improved mobile path prediction method is optimized through multiple rounds of iterative scenario simulation to obtain the mobile path prediction method that can adapt to various complex situations and dynamically respond to real-time environmental changes.

[0011] Optionally, based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure for the non-human biological invader is generated by an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged, including: Based on the trigger signal from the optimal alert response combination, using an intelligence-driven decision-making framework, analyzing and processing the real-time environmental interaction information of the trigger signal to obtain confirmation information that a non-human biological intruder has illegally crossed the perimeter; Based on the confirmed information, the selection and working mode of the eco-friendly interference device are evaluated, and the working mode with harmless stimulation that has been tested is selected to generate a setting plan for the repelling device; Using the repelling device setting scheme, the selected eco-friendly interference device is activated, and the repelling operation is performed according to the selected working mode to generate preliminary harmless repelling measures for the non-human biological intruder; Based on the preliminary harmless expulsion measures, feedback information collected from the real-time environmental interaction information of the non-human biological intruder is used to monitor the expulsion response of the non-human biological intruder to obtain a moving path tracking result of the non-human biological intruder; Based on the mobile path tracking results, the working mode of the eco-friendly interference device is dynamically adjusted to ensure that the eco-friendly interference device adapts to the behavioral characteristics of different types of non-human biological invaders to generate harmless expulsion measures, which are used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged.

[0012] Optionally, based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern is learned by using the anonymized real-time environmental interaction information, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to improve the sensitivity of the perimeter intrusion alarm, including: Based on the optimal warning response combination and the actual effect of the harmless expulsion measures, the group behavior pattern of non-human biological intruders is studied by using the anonymized real-time environmental interaction information to obtain the behavioral characteristics of the non-human biological intruders; According to the behavior characteristics, the group behavior pattern of the non-human biological invaders is analyzed to generate a behavior model for describing the behavior characteristics; Based on the behavior model, combined with the intelligence-driven decision-making framework and the natural language processing framework, the response strategy of the optimal alert response combination is evaluated and analyzed to obtain an improvement plan for the response strategy; Based on the improvement plan of the response strategy, the identified improvement points are optimized, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to ensure that the harmless expulsion measures are adaptable to different types of non-human biological intruders to improve the sensitivity of the perimeter intrusion alarm.

[0013] Optionally, based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure for the non-human biological invader is generated by an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged, including: Based on the optimal combination of alert responses When a non-human biological invader invades, a harmless expulsion measure is generated by an eco-friendly interference device to implement the non-human biological invader. , using the evaluation function Evaluate the harmless removal measures The effectiveness of the evaluation function is combined with the ecological impact score , expulsion effect score , resource consumption score and false positive rate penalty ; Evaluation function calculation formula: ; in, Represents the ecological impact score, which is used to judge the impact of the harmless removal measures on the ecosystem. represents the expulsion effect score, indicating the effective containment capability of the harmless expulsion measures on the non-human biological invaders. represents the resource consumption score, and FPR represents the false alarm rate penalty, which is used to reduce the proportion of non-intrusion events that are mistakenly identified as intrusions. is the weight coefficient, and , is the exponential factor, the weight coefficient and exponential factor The specific expression is: The specific expression of the exponential factor is: , , , , represents the frequency of ecological events in historical data, Indicates the dynamic change rate of the current environment. Indicates the load level, represents the false alarm rate, Indicates the corresponding variable value, is the adjustment coefficient, which is adjusted according to the actual situation; Based on the evaluation results of the harmless removal measures , using the eco-friendly jammer Get the harmless removal measures , ensuring that the harmless expulsion measures can effectively expel the non-human biological invaders without causing negative impacts on the ecosystem, Based on the evaluation function The output results are used to select the optimal harmless expulsion measures ; Calculation formula for harmless expulsion measures: ; in ， is the set of harmless removal measures, is the harmless removal measure mentioned above, Indicates that among all harmless expulsion measures, the evaluation function is selected The maximum harmless expulsion measure is the optimal harmless expulsion measure , is the optimal alert response combination, is the eco-friendly jammer.

[0014] Optionally, based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, and the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm, including: Based on the moving path estimation result , when an unusual activity pattern is detected The multimodal alert response mechanism is immediately activated when , evaluate the alert response combination The effectiveness of the evaluation function includes a safety score , timeliness score , cost score , and dynamic response score and false positive rate penalty ; Evaluation function calculation formula: ; in, Represents the safety score, Represents the timeliness score, represents the cost score, represents the dynamic response score, which measures the adaptability of the alert response combination to the behavior of the potential intruder under different time and space conditions. Represents the false alarm rate penalty, which is used to reduce the impact of false alarms. is the weight coefficient, and , It is an exponential factor used to adjust the impact of each scoring item on the overall evaluation. and exponential factor The specific expression is: , , , , , , Indicates the frequency of security events in historical data, Indicates the dynamic change rate of the current environment. Indicates the load level, represents the visibility of the potential intruder, represents the false alarm rate, Indicates the corresponding variable value, is the adjustment coefficient; Based on the moving path estimation result and unusual activity patterns , combined with an intelligence-driven decision-making framework and natural language processing technology Analyze the on-site status of the potential intruder , using genetic algorithms to search for each of the warning response combinations, and combining with Bayesian networks Perform scenario simulation to generate the evaluation function Maximize the best combination of alert responses; The calculation formula for the on-site status of the potential intruder is: ; in, is the moving path estimation result, is the abnormal activity pattern, is the result of the analysis by the natural language processing technique, is the information provided by the intelligence-driven decision-making framework, is the on-site status of the potential intruder, and the weight coefficient , adjust according to actual conditions to ensure that the impact of each factor on the overall assessment is reasonable, and In a second aspect, an embodiment of the present application provides a perimeter intrusion alarm system, including: A recognition module receives real-time environment interaction information from a plurality of distributed intelligent sensing nodes, identifies image features in the real-time environment interaction information through a convolutional neural network, and analyzes time series behavior patterns in the real-time environment interaction information to obtain a behavior pattern recognition result of the real-time environment interaction information; An adjustment module dynamically constructs a movement path of a potential intruder in a virtual three-dimensional geographic information system according to the behavior pattern recognition result, and adjusts the movement path prediction method in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; An analysis module, based on the movement path estimation result, immediately activates a multimodal alert response mechanism when an abnormal activity pattern is detected, analyzes the on-site status of the potential intruder using an intelligence-driven decision framework and natural language processing technology, and obtains an optimal alert response combination through a genetic algorithm; A generation module, based on the optimal warning response combination, generates harmless expulsion measures for the non-human biological intruders through an eco-friendly interference device when the non-human biological intruders invade, wherein the harmless expulsion measures are used to prevent the perimeter from being illegally crossed and to protect the ecological environment from being damaged; The optimization module uses the anonymized real-time environmental interaction information to learn group behavior patterns based on the actual effects of the optimal warning response combination and the harmless expulsion measures, and continuously optimizes the optimal warning response combination and the harmless expulsion measures to improve the sensitivity of the perimeter intrusion alarm.

[0015] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a perimeter intrusion alarm method as described in the first aspect above.

[0016] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a perimeter intrusion alarm method as described in the first aspect.

[0017] In an embodiment of the present application, real-time environmental interaction information is received from a plurality of distributed intelligent sensing nodes, image features in the real-time environmental interaction information are identified through a convolutional neural network, and time series behavior patterns in the real-time environmental interaction information are parsed to obtain a behavior pattern recognition result of the real-time environmental interaction information; based on the behavior pattern recognition result, a movement path of a potential intruder is dynamically constructed in a virtual three-dimensional geographic information system, and a prediction method of the movement path is adjusted in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; based on the movement path estimation result, a multimodal alert response mechanism is immediately activated when an abnormal activity pattern is detected, and intelligence is used to drive decision making The framework and natural language processing technology are used to analyze the on-site conditions of the potential intruder, and the optimal warning response combination is obtained through a genetic algorithm; based on the optimal warning response combination, when a non-human biological intruder invades, a harmless expulsion measure for the non-human biological intruder is generated through an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged; based on the actual effects of the optimal warning response combination and the harmless expulsion measure, the real-time environmental interaction information after anonymization is used to learn the group behavior pattern, and the optimal warning response combination and the harmless expulsion measure are continuously optimized to improve the sensitivity of the perimeter intrusion alarm.

[0018] The technical solution of this application has the following beneficial effects: This application ensures high-precision detection and behavior pattern recognition of potential intruders through multi-source information fusion and intelligent perception technology. By recognizing image features through convolutional neural networks and parsing time series behavior patterns, the system can accurately capture environmental changes. The mobile path is dynamically constructed using a virtual three-dimensional geographic information system, and the prediction method is adjusted in combination with Bayesian optimization and reinforcement learning algorithms to ensure the accuracy and real-time nature of the invasion path estimation. At the same time, eco-friendly and harmless expulsion measures are introduced to ensure that the ecological environment is not damaged while protecting safety.

[0019] Furthermore, anomaly detection and time series prediction analysis are used to ensure that the system can quickly generate immediate response measures for abnormal activity patterns. Based on the estimated results of the movement path, the multimodal alert response mechanism is immediately activated. Combining the intelligence-driven decision-making framework with natural language processing technology, graph neural networks and knowledge graphs are used to analyze the on-site conditions of potential intruders to ensure the generation of detailed assessment reports. The best alert response combination is searched through genetic algorithms and particle swarm optimization, and continuous adjustment and optimization is made using reinforcement learning strategies to ensure that the ability to respond to emergencies and the response efficiency are significantly improved.

[0020] Furthermore, through the intelligence-driven decision-making framework and natural language processing technology, the multi-source intelligence information collected in the immediate response measures is integrated and analyzed to ensure the generation of a field situation analysis report including perimeter intrusion environment information and real-time monitoring information. The unstructured multi-source intelligence information is converted into structured description information through natural language processing technology, and the knowledge graph of potential intruders is updated using graph neural networks to ensure the provision of a comprehensive assessment report. Based on the entity relationships in the updated knowledge graph, the identity, capabilities, and behavior trajectories of potential intruders are identified and analyzed to ensure that a scientific basis is provided for subsequent security decisions, enhancing the intelligence and adaptability of the system.

[0021] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A flow chart of a perimeter intrusion alarm method provided by the present application is shown; Figure 2 A schematic diagram of the structure of a perimeter intrusion alarm system provided by the present application is shown; Figure 3 A schematic diagram of the structure of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0025] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0027] Figure 1 A flowchart of a perimeter intrusion alarm method is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes: 101. Receive real-time environment interaction information from multiple distributed intelligent perception nodes, identify image features in the real-time environment interaction information through a convolutional neural network, and parse time series behavior patterns in the real-time environment interaction information to obtain a behavior pattern recognition result of the real-time environment interaction information; In this step, real-time environmental interaction information is received from multiple distributed intelligent sensing nodes, which may include cameras, infrared sensors, sound sensors, etc. Real-time environmental interaction information includes multiple types of data such as image data, temperature changes, sound signals, etc., which are used to comprehensively capture the dynamic situation inside and outside the perimeter. By recognizing image features through convolutional neural networks and parsing time series behavior patterns, the system can extract meaningful behavior pattern recognition results from a large amount of data. This technology not only improves detection accuracy, but also captures the activity patterns of potential intruders in complex environments.

[0028] In an embodiment of the application, in a large industrial park, cameras and sensors installed on the perimeter continuously monitor the surrounding environment. When an unidentified object approaches, the camera captures its image, the infrared sensor records its temperature change, and the sound sensor collects the sound generated by its movement. These data are transmitted to the central processing unit in real time, and the convolutional neural network is used to identify the object as a human or non-human creature and analyze its behavior pattern. For example, if a human is detected moving quickly and approaching the perimeter, the system will mark this behavior as a potential threat and prepare to proceed to the next step.

[0029] 102. According to the behavior pattern recognition result, dynamically construct the movement path of the potential intruder in the virtual three-dimensional geographic information system, and adjust the prediction method of the movement path in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; In this step, the movement path of potential intruders is dynamically constructed through the virtual 3D GIS, which provides a realistic geospatial model that enables the system to simulate and predict the movement trajectory of intruders in a virtual environment. The prediction method of the movement path is adjusted by combining Bayesian optimization and reinforcement learning algorithms to ensure the accuracy and real-time nature of the prediction results. This step helps to formulate response strategies in advance and improve the system's response speed.

[0030] In the application embodiment, continuing with the above industrial park example, once the central processing unit confirms the existence of a potential threat, the system immediately constructs the person's movement path in the virtual three-dimensional geographic information system. By selecting the most likely path through Bayesian optimization, the system predicts that the person will continue to move in a straight line toward a specific area. As further data is collected, such as new camera images and sensor readings, the reinforcement learning algorithm continuously adjusts the path prediction to ensure that the estimated results are consistent with the actual behavior. At this point, the system is ready to activate the multimodal alert response mechanism.

[0031] 103. Based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm; In this step, when an abnormal activity pattern is detected, the system immediately activates the multimodal alert response mechanism. The multimodal alert response mechanism combines multiple types of alert measures, such as sound and light alarms, video surveillance, drone patrols, etc. Using the intelligence-driven decision-making framework and natural language processing technology, the system can analyze the on-site conditions of potential intruders and generate detailed assessment reports. Genetic algorithms are used to search for the best alert response combination to ensure the most effective response measures.

[0032] In the application embodiment, in the industrial park scenario, when the system predicts that someone is approaching the perimeter and behaving abnormally, the multimodal alert response mechanism is immediately activated. The sound and light alarm issues a warning to remind nearby people to pay attention to safety; the video surveillance system switches to high-resolution mode to record the behavior of the intruder in detail; the drone takes off for aerial patrol to obtain a wider field of view. At the same time, the intelligence-driven decision-making framework analyzes all sensor data and generates an assessment report containing information such as the identity and movement trajectory of the intruder. The genetic algorithm searches for the best alert response combination, such as increasing the patrol frequency and deploying more monitoring equipment to ensure that potential threats can be effectively prevented.

[0033] 104. Based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure for the non-human biological invader is generated through an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged; In this step, the system uses eco-friendly interference devices to implement harmless expulsion measures for non-human biological invaders. These measures are designed to prevent illegal crossing of the perimeter while protecting the ecological environment from damage. Through scientifically designed sound waves, light waves or other environmentally friendly means, the system can effectively drive away animals without causing harm to them. This approach not only protects the integrity of the perimeter, but also reflects a sense of responsibility for the ecological environment.

[0034] In the application embodiment, in the implementation of the industrial park, when the system identifies a fox trying to cross the perimeter, the eco-friendly jammer is immediately activated. According to the behavior pattern of the fox, the system chooses to drive it away by simulating the call of the natural enemy. After hearing the call, the fox quickly changes direction and moves away from the perimeter area. This process not only avoids harm to the fox, but also ensures the integrity and safety of the perimeter. The system records this incident to provide data support for further optimization of the expulsion measures.

[0035] 105. Based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern is learned by using the anonymized real-time environmental interaction information, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to improve the sensitivity of the perimeter intrusion alarm.

[0036] In this step, based on the actual effects of the best alert response combination and harmless expulsion measures, the system uses anonymized real-time environmental interaction information to learn group behavior patterns. Through machine learning algorithms, the system continuously optimizes alert response combinations and harmless expulsion measures to improve the sensitivity and effectiveness of perimeter intrusion alarms. This process is continuous, ensuring that the system can adapt to new threats and environmental changes and maintain long-term efficient performance.

[0037] In the application embodiment, during the long-term operation of the industrial park, the system continuously collects and analyzes data on various types of intrusion incidents, including activity patterns of humans and non-human organisms. By learning from these data, the system gradually optimizes the warning response combination and harmless expulsion measures. For example, in response to frequent bird invasions, the system adjusted the operating frequency and intensity of the sonic expulsion device, significantly reducing the number of times birds approach the perimeter. Similarly, for human intruders, the system improves the multimodal warning response mechanism, making the response faster and more accurate. Through this continuous optimization, the security protection level of the industrial park has been comprehensively improved.

[0038] In summary, through the above steps 101 to 105, the perimeter intrusion alarm method realizes the intelligent perception and analysis of real-time environmental interaction information, the dynamic construction and prediction of the potential intruder's movement path, and the instant activation and optimization of the multi-modal warning response mechanism, and finally forms a set of efficient, flexible and environmentally friendly security protection system. This series of steps complement each other and jointly build an intelligent and highly reliable perimeter security protection system suitable for complex and changing application scenarios.

[0039] Optionally, in step 103, based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm, including: Based on the movement path prediction result, the movement path prediction result is analyzed by using anomaly detection and time series prediction. When an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated to generate immediate response measures for the abnormal activity pattern; according to the immediate response measures, in combination with the intelligence-driven decision-making framework and natural language processing technology, a graph neural network and a knowledge graph are used to parse the on-site conditions of the potential intruder to obtain an assessment report of the potential intruder; based on the assessment report of the potential intruder, a corresponding alert response combination is searched by using a genetic algorithm and a particle swarm optimization, and a reinforcement learning strategy is used to continuously adjust and optimize according to real-time feedback information of the actual situation of the potential intruder to generate an optimal alert response combination.

[0040] Optionally, according to the immediate response measures, in combination with the intelligence-driven decision-making framework and natural language processing technology, the on-site conditions of the potential intruder are analyzed using a graph neural network and a knowledge graph to obtain an assessment report of the potential intruder, including: According to the immediate response measures, the multi-source intelligence information on the potential intruder collected in the immediate response measures is integrated and analyzed using an intelligence-driven decision-making framework to generate an on-site situation analysis report including perimeter intrusion environment information and real-time monitoring information; based on the on-site situation analysis report, the unstructured multi-source intelligence information in the on-site situation analysis report is analyzed using natural language processing technology, and key information of the potential intruder is extracted to generate structured description information of the key information; based on the structured description information, the knowledge graph of the potential intruder in the structured description information is updated using a graph neural network to obtain an updated knowledge graph; based on the entity relationships in the updated knowledge graph, the on-site situation including the identity, capabilities and behavior trajectory of the potential intruder is identified and parsed to obtain an evaluation report on the potential intruder.

[0041] In this step, the estimated movement paths are analyzed through anomaly detection and time series prediction to ensure that the system can quickly identify and respond to abnormal activity patterns. Anomaly detection includes real-time monitoring based on statistical models or machine learning algorithms to detect signs of deviation from normal behavior; time series prediction uses historical data to predict future trends. The intelligence-driven decision-making framework integrates multi-source intelligence information, such as video, audio, and sensor data, to generate on-site situation analysis reports. Natural language processing technology parses unstructured information, extracts key details, and forms structured descriptions. Graph neural networks and knowledge graphs further update the identity, capabilities, and behavior trajectories of potential intruders to ensure that the assessment report is comprehensive and accurate. Genetic algorithms and particle swarm optimization search for the best warning response combination, and reinforcement learning strategies continuously adjust and optimize based on real-time feedback to ensure the effectiveness and flexibility of response measures.

[0042] In an embodiment of the present application, when the system detects an abnormal activity pattern, the multimodal alert response mechanism is immediately activated, and the estimated results of the movement path are analyzed using anomaly detection and time series prediction to generate immediate response measures. Subsequently, the intelligence-driven decision-making framework is combined with natural language processing technology, and the graph neural network and knowledge graph are used to analyze the on-site conditions of potential intruders and generate a detailed evaluation report. Specifically, the system first integrates and analyzes the multi-source intelligence information collected in the immediate response measures to generate an on-site condition analysis report containing environmental and real-time monitoring information. Then, the unstructured information is converted into a structured description through natural language processing technology, and the knowledge graph of potential intruders is updated. Finally, based on the entity relationships in the updated knowledge graph, the key features of potential intruders are identified and parsed to generate a final evaluation report. Genetic algorithms and particle swarm optimization search for the best alert response combination, and reinforcement learning strategies are used to continuously adjust and optimize according to real-time feedback to ensure the best effect of the response measures.

[0043] In the application scenario of the industrial park, when the system predicts that a human is approaching the perimeter quickly and is behaving abnormally, it immediately enters the optional solution of step 103. First, the system uses anomaly detection and time series prediction to confirm that the behavior is an abnormal activity pattern, and then activates a multimodal alert response mechanism, such as sound and light alarms and drone patrols. Next, the system integrates multi-source intelligence information such as camera images and infrared sensor data to generate a site situation analysis report containing environmental and real-time monitoring information. Natural language processing technology parses the unstructured information in the report and extracts key details, such as the intruder's route and possible purpose. The graph neural network updates the intruder's knowledge graph to determine his identity and intention. Based on this information, genetic algorithms and particle swarm optimization find the best alert response combination, such as increasing patrol frequency and deploying more monitoring equipment. At the same time, the system continuously optimizes the response measures based on real-time feedback through reinforcement learning strategies to ensure that each response is more accurate and effective. This series of operations not only improves the system's sensitivity, but also enhances its adaptive ability.

[0044] Optionally, in step 102, based on the behavior pattern recognition result, a movement path of a potential intruder is dynamically constructed in a virtual three-dimensional geographic information system, and a prediction method of the movement path is adjusted in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result, including: According to the behavior pattern recognition result, the behavior pattern recognition result is analyzed by using a virtual three-dimensional geographic information system to obtain a movement path model of each potential intruder; based on the movement path model, the movement path model is evaluated and optimized by using Bayesian optimization to generate a preliminary movement path prediction method to improve the accuracy of the movement path prediction of the potential intruder; based on the preliminary movement path prediction method, a reinforcement learning algorithm is subjected to scenario simulation analysis and processing, and the preliminary movement path prediction method is further adjusted to obtain the movement path prediction method; based on the movement path prediction method, the correction information of the prediction method fed back by reinforcement learning is used to analyze and correct the uncertainty and error in the movement path model to generate a movement path estimation result.

[0045] Optionally, based on the preliminary movement path prediction method, a scenario simulation analysis process is performed on the reinforcement learning algorithm, and the preliminary movement path prediction method is further adjusted to obtain the movement path prediction method, including: Based on the preliminary mobile path prediction method, the terrain conditions, weather conditions and emergency scenarios of the potential intruder's invasion are simulated to generate simulation scenarios including various mobile paths; according to the simulation scenarios, the reinforcement learning algorithm is applied to simulate the behavior pattern of the potential intruder, and the simulation results are compared and analyzed with the preset ideal behavior pattern of the potential intruder, the actual performance of the preliminary mobile path prediction method is evaluated, and an analysis report of the preliminary mobile path prediction method is obtained; based on the analysis report, the reward mechanism and the penalty mechanism in the reinforcement learning algorithm are adjusted to optimize the preliminary mobile path prediction method and generate an improved mobile path prediction method; based on the improved mobile path prediction method, the improved mobile path prediction method is optimized through multiple rounds of iterative scenario simulation to obtain the mobile path prediction method that can adapt to a variety of complex situations and dynamically respond to real-time environmental changes.

[0046] In this step, the behavior pattern recognition results are analyzed using a virtual 3D geographic information system to generate a movement path model for each potential intruder. These models include data such as historical trajectory, speed, and direction to predict future movement paths. Bayesian optimization evaluates and optimizes the path model to improve accuracy. The reinforcement learning algorithm adjusts the preliminary prediction method through scenario simulation analysis and processing to correct uncertainty and error. The final movement path estimation results can adapt to a variety of complex situations and dynamically respond to real-time environmental changes.

[0047] In the embodiment of the present application, the system first constructs a movement path model of each potential intruder in a virtual three-dimensional geographic information system based on the behavior pattern recognition results. Then, Bayesian optimization is used to evaluate and optimize these models to generate a preliminary movement path prediction method. Then, based on the preliminary prediction method, the reinforcement learning algorithm is simulated in scenarios, including terrain conditions, weather conditions, and emergencies. By comparing the simulation results with the ideal behavior pattern, the reward mechanism and penalty mechanism in the reinforcement learning are evaluated and adjusted. After multiple rounds of iterative optimization, a movement path prediction method that adapts to a variety of complex situations and dynamically responds to real-time environmental changes is generated.

[0048] In order to further improve the sensitivity of perimeter intrusion alarms, in the application of industrial parks, the system uses the behavioral pattern recognition results collected by cameras and sensors to build an intruder's movement path model in a virtual three-dimensional geographic information system. Bayesian optimization improves the accuracy of path prediction. Subsequently, the system simulated intrusion scenarios under different weather and terrain conditions and applied reinforcement learning algorithms to analyze the intruder's behavior patterns. After multiple rounds of iterative optimization, the system generated an improved movement path prediction method. When a human is detected approaching the perimeter quickly, the system immediately activates the alert response mechanism to ensure that the response measures are accurate and effective, and seamlessly connected with subsequent steps, improving the overall protection level.

[0049] Optionally, in step 104, based on the optimal warning response combination, when a non-human biological intruder invades, a harmless expulsion measure for the non-human biological intruder is generated by an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged, including: According to the trigger signal from the optimal warning response combination, the real-time environmental interaction information of the trigger signal is analyzed and processed using an intelligence-driven decision-making framework to obtain confirmation information that a non-human biological intruder has illegally crossed the perimeter; based on the confirmation information, the selection and working mode of the eco-friendly interference device are evaluated, and the tested working mode with harmless stimulation is selected to generate a repelling device setting plan; using the repelling device setting plan, the selected eco-friendly interference device is started, and the repelling operation is performed according to the selected working mode to generate an initial response to the non-human biological intruder. Based on the preliminary harmless expulsion measures, the feedback information collected from the real-time environmental interaction information of the non-human biological invaders is used to monitor the expulsion response of the non-human biological invaders to obtain the movement path tracking results of the non-human biological invaders; based on the movement path tracking results, the working mode of the eco-friendly interference device is dynamically adjusted to ensure that the eco-friendly interference device can adapt to the behavioral characteristics of different types of non-human biological invaders to generate harmless expulsion measures, which are used to prevent the perimeter from being illegally crossed and protect the ecological environment from being destroyed.

[0050] In this step, the real-time environmental interaction information of the trigger signal is analyzed using the intelligence-driven decision-making framework to confirm the illegal crossing behavior of non-human biological intruders. Based on this confirmation information, the system evaluates and selects the working mode of the eco-friendly interference device to ensure the effectiveness of harmless stimulation. The repelling device setting scheme guides the start of the selected device and performs preliminary repelling operations according to the predetermined working mode. By collecting feedback information in real time, monitoring the repelling response, generating movement path tracking results, and dynamically adjusting the device working mode to adapt to the behavioral characteristics of different types of intruders.

[0051] In the embodiment of the present application, when the optimal warning response combination is triggered, the system first analyzes the real-time environmental interaction information of the trigger signal to confirm the illegal crossing behavior of non-human biological intruders. Subsequently, based on the confirmation information, the tested eco-friendly interference device and its harmless stimulation working mode are evaluated and selected to generate a repelling device setting plan. After starting the selected device, the preliminary repelling operation is performed according to the set working mode. The system continuously collects feedback information, monitors the repelling reaction, generates the moving path tracking results, and dynamically adjusts the device working mode according to the results to ensure that the repelling measures are effective and environmentally friendly.

[0052] In the application of industrial parks, when step 102 predicts that a fox is approaching the perimeter, the system activates the multimodal warning response mechanism through step 103. Once it is confirmed that the fox is trying to cross illegally, the system selects an eco-friendly jammer that simulates the call of a natural enemy. After starting the device, the system monitors the reaction of the fox through cameras and sensors and generates movement path tracking results. If the fox is not effectively driven away, the system dynamically adjusts the working mode of the device, such as changing the frequency or intensity, until it is successfully driven away. This process is closely connected with the previous steps to ensure the integrity of the perimeter and the safety of the ecological environment.

[0053] Optionally, in step 105, based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern is learned by using the anonymized real-time environmental interaction information, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to improve the sensitivity of the perimeter intrusion alarm, including: Based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern of non-human biological intruders is learned by using the real-time environmental interaction information after anonymization to obtain the behavioral characteristics of the non-human biological intruders; based on the behavioral characteristics, the group behavior pattern of the non-human biological intruders is analyzed to generate a behavioral model for describing the behavioral characteristics; based on the behavioral model, combined with the intelligence-driven decision-making framework and the natural language processing framework, the response strategy of the optimal warning response combination is evaluated and analyzed to obtain an improved solution for the response strategy; based on the improved solution for the response strategy, the identified improvement points are optimized, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to ensure that the harmless expulsion measures are adapted to different types of non-human biological intruders to improve the sensitivity of the perimeter intrusion alarm.

[0054] In this step, the group behavior patterns of non-human biological intruders are learned using anonymized real-time environmental interaction information. These data include sensor records, camera images, etc., which are used to analyze the behavioral characteristics of intruders. By generating a behavioral model that describes the behavioral characteristics, the system combines the intelligence-driven decision-making framework and natural language processing technology to evaluate and optimize the response strategy of the best warning response combination. Finally, the system continuously optimizes the harmless expulsion measures according to the improvement plan to ensure that they are suitable for different types of non-human biological intruders and improve the sensitivity of perimeter intrusion alarms.

[0055] In the embodiment of the present application, based on the actual effects of the optimal warning response combination and harmless expulsion measures, the system uses the real-time environmental interaction information after anonymization to learn the group behavior patterns of non-human biological invaders and extract their behavioral characteristics. Then, the system generates a behavioral model that describes these characteristics, and combines the intelligence-driven decision-making framework and natural language processing technology to evaluate the effectiveness of the existing response strategy. Based on the evaluation results, the improvement points are identified and optimized, and the warning response combination and harmless expulsion measures are continuously adjusted to adapt to the behavioral characteristics of different types of invaders, thereby improving the sensitivity and response efficiency of the system.

[0056] In the application of the industrial park, the system predicts that foxes frequently approach the perimeter through step 102, activates the multimodal alert response mechanism in step 103, and successfully implements eco-friendly expulsion measures in step 104. Subsequently, in step 105, the system uses the real-time environmental interaction information after anonymization to learn the behavior patterns of the fox group and generate a behavior model. Combining the intelligence-driven decision-making framework and natural language processing technology, the system evaluates the effectiveness of the current response strategy and finds that the sound wave frequency needs to be adjusted to more effectively drive away foxes. Based on this, the system optimizes the expulsion measures to ensure that different types of non-human biological intruders can be effectively responded to, thereby improving the sensitivity and reliability of the overall intrusion alarm system.

[0057] This application takes into account that traditional expulsion measures often ignore the impact on the ecological environment, which may lead to unnecessary ecological damage or waste of resources. At the same time, the system has a high false alarm rate, which reduces the reliability and efficiency in practical applications. By introducing an evaluation function, combining the ecological impact score, expulsion effect score, resource consumption score and false alarm rate penalty, the effectiveness of harmless expulsion measures can be comprehensively evaluated to ensure that the measures are both effective and environmentally friendly. Therefore, a new optional solution is proposed, which includes: Optionally, based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure for the non-human biological invader is generated by an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged, including: Based on the optimal combination of alert responses When a non-human biological invader invades, a harmless expulsion measure is generated by an eco-friendly interference device to implement the non-human biological invader. , using the evaluation function Evaluate the harmless removal measures The effectiveness of the evaluation function is combined with the ecological impact score , expulsion effect score , resource consumption score and false positive rate penalty ; Evaluation function calculation formula: ; in, Represents the ecological impact score, which is used to judge the impact of the harmless removal measures on the ecosystem. represents the expulsion effect score, indicating the effective containment capability of the harmless expulsion measures on the non-human biological invaders. represents the resource consumption score, and FPR represents the false alarm rate penalty, which is used to reduce the proportion of non-intrusion events that are mistakenly identified as intrusions. is the weight coefficient, and , is the exponential factor, the weight coefficient and exponential factor The specific expression is: , the specific expression of the exponential factor is: , , , , represents the frequency of ecological events in historical data, Indicates the dynamic change rate of the current environment. Indicates the load level, represents the false alarm rate, Indicates the corresponding variable value, It is the adjustment coefficient, which is adjusted according to the actual situation; Based on the evaluation results of the harmless removal measures , using the eco-friendly jammer Get the harmless removal measures , ensuring that the harmless expulsion measures can effectively expel the non-human biological invaders without causing negative impacts on the ecosystem, Based on the evaluation function The output results are used to select the optimal harmless expulsion measures ; Calculation formula for harmless expulsion measures: ; in ， is the set of harmless removal measures, is the harmless removal measure mentioned above, Indicates that among all harmless expulsion measures, the evaluation function is selected The maximum harmless expulsion measure is the optimal harmless expulsion measure , is the optimal alert response combination, It is the eco-friendly jammer mentioned above; the overall design of the formula is to comprehensively evaluate the effectiveness of harmless expulsion measures, combining four factors: ecological impact, expulsion effect, resource consumption and false alarm rate. By optimizing the weight coefficient and exponential factor, it ensures that the measures can effectively expel non-human biological intruders while minimizing the impact on the ecological environment, thereby improving the sensitivity and reliability of the perimeter intrusion alarm system.

[0058] The following is a brief explanation of the design reasons for each item in the formula: The design reason of the evaluation function is to comprehensively consider the multi-faceted impacts of harmless repellent measures, including ecological impact, repellent effect, resource consumption and false alarm rate. By introducing weight coefficients and exponential factors, the importance of each factor is accurately quantified to ensure that the repellent measures can effectively curb invaders while minimizing the negative impact on the ecological environment, thereby improving the overall performance and reliability of the system.

[0059] The formula of the evaluation function is: :This sub-item emphasizes the friendliness of harmless expulsion measures to the ecosystem by introducing weights and exponential factor , ensuring that ecological impacts are fully considered during the assessment process, with higher and appropriate The value can improve the overall assessment score and encourage the use of expulsion measures with the least impact on the ecosystem; :This sub-item focuses on the effectiveness of the expulsion effect, with a weight of and exponential factor Ensure that the actual effectiveness of the removal measures is taken into account in the evaluation, and the higher and appropriate The value can significantly improve the evaluation score and promote the selection of efficient expulsion methods; :This item considers the cost-effectiveness of resource consumption and has a negative weight and exponential factor Emphasize the importance of reducing resource consumption, lower and appropriate The value helps to reduce the overall evaluation score and encourages the selection of expulsion measures with more optimal resource utilization; :This item aims to reduce the false positive rate, negative weight and exponential factor Emphasizing the importance of reducing false positives, lower and appropriate The value can significantly reduce the evaluation score and promote the selection of expulsion measures with lower false positive rates; The purpose of adding up the sub-items is to comprehensively consider multiple evaluation dimensions and ensure that the final evaluation results can fully reflect the multi-faceted performance of harmless removal measures. Each sub-item has a clear physical meaning and weight. By adding them together, a comprehensive score can be obtained to balance the relationship between different factors. :Indicates harmless removal measures A comprehensive evaluation score to measure its effectiveness; They represent the weights of ecological impact score, expulsion effect score, resource consumption score and false alarm rate penalty respectively. These weights are calculated to ensure that the sum is 1, and the importance of each factor is adjusted according to the actual situation; : Evaluate the impact of harmless expulsion measures on the ecosystem. The higher the value, the smaller the negative impact on the ecosystem. : Expulsion effect score, which measures the effectiveness of harmless expulsion measures in deterring non-human biological invaders. The higher the value, the better the expulsion effect. : Resource consumption score, which evaluates the resource consumption required for harmless expulsion measures. The higher the value, the lower the resource consumption. : False alarm rate penalty, which measures the proportion of non-intrusion events that the system mistakenly identifies as intrusions. The higher the value, the lower the false alarm rate; : Exponential factor, used to adjust the sensitivity of each scoring item. The exponential factor is calculated based on parameters such as the frequency of ecological events in historical data, the dynamic change rate of the current environment, the load level and the false alarm rate to adapt to different application scenarios.

[0060] The reason for designing the harmless expulsion measure calculation formula is to combine the best warning response combination, ecological impact, expulsion effect and environmental change prediction, and optimize the expulsion strategy through time series model and reinforcement learning algorithm. Ensure that the measures can effectively expel non-human biological invaders without causing negative impact on the ecosystem, and improve the adaptability and efficiency of the perimeter security system.

[0061] The formula for harmless expulsion measures is: : Harmless expulsion measures, which represent the final harmless expulsion strategy, aiming to effectively expel non-human biological invaders while minimizing the impact on the ecosystem; : Eco-friendly jammers refer to specific equipment or methods used to implement harmless expulsion, such as simulated calls of natural enemies, flashing lights, sound waves of specific frequencies, etc. These devices are designed and tested to ensure that they will not have a negative impact on the environment while repelling intruders; : The best alert response combination is based on the best response strategy obtained by the multimodal alert response mechanism and genetic algorithm. This combination takes into account multiple factors, such as the type of intruder, behavior pattern, geographical location, etc., to ensure the most effective alert and response measures; : The best harmless expulsion measure; : A collection of all possible harmless repellent measures, such as different settings or combinations of eco-friendly distraction devices such as different sound frequencies, light intensity changes, scent release, etc.

[0062] The following is a brief explanation of how to obtain the parameters of the formula: Weight coefficient :Through calculation, based on historical data and the dynamic changes of the current environment, these coefficients are adjusted according to the actual situation to ensure that the importance of each factor is reasonably reflected; Ecological Impact Rating : By analyzing the impact of harmless expulsion measures on the ecosystem, combined with the frequency of ecological events in historical data Based on the real-time monitoring information, a high score indicates a small negative impact on the ecosystem, while a low score indicates the opposite, which can be obtained through expert evaluation or long-term monitoring data statistics; Repellent effect rating :Based on the feedback information of actual expulsion effect and the results of simulation analysis, a high score indicates a good expulsion effect, and a low score indicates a poor effect. The data of actual expulsion effect can be collected through field tests, sensor data and video surveillance records; Resource Consumption Rating : Evaluate the resource consumption required for harmless expulsion measures, including energy, materials, etc. The higher the value, the lower the resource consumption, which can be obtained through equipment operation logs, energy consumption records and cost accounting; False Positive Rate Penalty : Measures the proportion of non-intrusion events that the system mistakenly identifies as intrusions. The higher the value, the lower the false alarm rate. The false alarm rate can be calculated through system logs, user feedback, and alarm records. Exponential Factor :Based on the frequency of ecological events in historical data , the dynamic change rate of the current environment , load level and false positive rate Calculated; Safe removal measures : By comprehensively considering the best warning response combination, ecological impact score, expulsion effect score, time series prediction model and reinforcement learning algorithm results, we finally select the harmless expulsion strategy that has been tested and proven to be effective; Eco-friendly jammer : Choose repellent devices that have been tested and proven to be harmless to the ecosystem, such as simulated calls of natural enemies, flashing lights, sound waves of specific frequencies, etc. The selection of these devices is based on laboratory testing and field application verification; The best combination of alert and response : Optimized by genetic algorithm and multimodal alert response mechanism, this combination takes into account multiple factors, such as the type of intruder, behavior pattern, geographic location, etc., to ensure the most effective alert and response measures are taken, and its effectiveness is verified through simulation and actual application; Ecological Impact Rating :Same as in the evaluation function formula The acquisition method is consistent, and is obtained by analyzing the impact of harmless expulsion measures on the ecosystem, combining historical data and real-time monitoring information; Repellent effect rating :Same as in the evaluation function formula The acquisition method is consistent, based on the feedback information of the actual expulsion effect and the simulation analysis results; Time Series Forecasting Models : Using time series analysis technology, we predict future environmental changes based on historical data and real-time monitoring information. The model is trained through machine learning algorithms and can adapt to environmental changes in different scenarios to improve long-term effectiveness; : The optimal harmless expulsion measure is calculated according to the formula; : A collection of all possible harmless repellent measures, such as different sound frequencies, light intensity changes, scent release and other eco-friendly distraction devices, calculated based on different settings or combinations; Ecological impact scoring: Use sensor networks to collect environmental parameters (such as air quality, water quality, noise level, etc.) and compare them with historical baseline values; evaluate changes in biodiversity in the area through field surveys or remote monitoring; invite ecologists and environmental experts to score based on actual conditions. Scoring range: 0 (worst) to 1 (best); Repellent effectiveness scoring: Test the effects of different repellent measures on target organisms in a controlled environment and record the success rate; apply repellent measures in a real environment and track the reactions of intruders through monitoring equipment; collect user feedback after implementing repellent measures to understand their effectiveness. Scoring range: 0 (ineffective) to 1 (completely effective); Resource consumption score: directly measures the amount of electricity, fuel, and other resources used; checks the lifespan and maintenance needs of equipment; calculates the cost of each eviction operation, including equipment procurement, maintenance, and operating expenses. Scoring range: 0 (extremely high consumption) to 1 (extremely low consumption); Safety scoring: Use quantitative and qualitative risk assessment tools to assess potential safety hazards; refer to historical accident data in similar scenarios to predict the safety of the current solution; ensure that all measures comply with relevant laws, regulations and industry standards. Scoring range: 0 (unsafe) to 1 (very safe); Timeliness score: Record the time interval from detecting the intruder to initiating the expulsion measures; evaluate the effective duration of the expulsion measures to ensure that it is long enough to prevent the intrusion; obtain immediate feedback through the real-time monitoring system and adjust the response speed. The score range is from 0 (extremely slow) to 1 (very fast); Cost scoring: Calculate the construction, operation and maintenance costs of the entire expulsion system; compare the cost of investment with the expected benefits and evaluate the economic benefits; consider future maintenance and upgrade costs and make long-term cost forecasts. Scoring range: 0 (very high cost) to 1 (very low cost); Dynamic response score: Tests the adaptability of the expulsion measures under different environmental conditions, such as weather changes, terrain differences, etc.; evaluates whether the system can quickly adjust its strategy to respond to new threats or changes; examines whether the system has the ability to self-optimize and learn and improve from past experience. Scoring range: 0 (fixed) to 1 (highly flexible); Reinforcement Learning Algorithms : By continuously collecting and analyzing the feedback information of the actual expulsion effect, the expulsion strategy is dynamically adjusted to ensure that it adapts to different types of intruders and complex and changing environmental conditions. The algorithm is optimized through multiple iterations to continuously improve the effectiveness and reliability of the expulsion measures.

[0063] In the embodiment of the present application, in a large industrial park, the system detects that foxes frequently approach the perimeter. Based on the prediction results and the alert response mechanism, the system decides to use an eco-friendly jammer to drive away foxes. In order to ensure the effectiveness and environmental friendliness of the drive-away measures, the system uses the above formula to evaluate and optimize the drive-away strategy.

[0064] Parameter settings: Frequency of ecological events in historical data: 0.2 (frequency of similar events in the past year); The dynamic change rate of the current environment (D): 0.5 (the current environment change rate); Load level (L): 0.8 (current system load level); False alarm rate (F): 0.1 (false alarm rate of the current system); Adjustment coefficients: ; Time series forecasting model: predict environmental changes in the next week; Reinforcement learning algorithm: continuously adjust and optimize expulsion measures; Calculation process: Calculate weight coefficient : ; Calculate the exponential factor : ; Calculate the evaluation function : Assumptions ;; ; Evaluation Function The calculation result is 0.134, indicating that the harmless expulsion measures under the current configuration have a good overall effect when considering ecological impact, expulsion effect, resource consumption and false alarm rate. Specifically: A higher Ecological Impact Index (EI) means that the removal measure has less impact on the ecosystem.

[0065] A high expulsion effectiveness score (DE) indicates that the expulsion measures have a strong ability to effectively deter foxes.

[0066] The resource consumption score (RC) and false positive rate penalty (FPR) are low, reducing the possibility of resource waste and false positives.

[0067] Based on this evaluation result, the system chooses to activate an eco-friendly jammer that simulates the calls of natural enemies, and optimizes future expulsion strategies based on the time series prediction model. At the same time, the reinforcement learning algorithm is used to continuously adjust and optimize the expulsion measures to ensure long-term and efficient operation. This series of operations not only improves the sensitivity and reliability of the system, but also protects the ecological environment and achieves harmonious coexistence between man and nature.

[0068] This application considers that traditional methods often rely on fixed rules and are difficult to adapt to different environmental changes and intrusion behavior patterns. By introducing an evaluation function that combines security score, timeliness score, cost score, dynamic response score and false alarm rate penalty, the effectiveness of the alert response combination can be comprehensively evaluated. Therefore, a new optional solution is proposed, which includes: Optionally, based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, and the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm, including: Based on the moving path estimation result , when an unusual activity pattern is detected The multimodal alert response mechanism is immediately activated when , evaluate the alert response combination The effectiveness of the evaluation function includes a safety score , timeliness score , cost score , and dynamic response score and false positive rate penalty ; Evaluation function calculation formula: ; in, Represents the safety score, Represents the timeliness score, represents the cost score, represents the dynamic response score, which measures the adaptability of the alert response combination to the behavior of the potential intruder under different time and space conditions. Represents the false alarm rate penalty, which is used to reduce the impact of false alarms. is the weight coefficient, and , It is an exponential factor used to adjust the impact of each scoring item on the overall evaluation. and exponential factor The specific expression is: , , , , , , Indicates the frequency of security events in historical data, Indicates the dynamic change rate of the current environment. Indicates the load level, represents the visibility of the potential intruder, represents the false alarm rate, Indicates the corresponding variable value, is the adjustment coefficient; Based on the moving path estimation result and unusual activity patterns , combined with an intelligence-driven decision-making framework and natural language processing technology Analyze the on-site status of the potential intruder , using genetic algorithm Search for each of the alert response combinations and combine with the Bayesian network Perform scenario simulation to generate the evaluation function Maximize the best combination of alert responses; The calculation formula for the on-site status of the potential intruder is: ; in, is the moving path estimation result, is the abnormal activity pattern, is the result of the analysis by the natural language processing technique, is the information provided by the intelligence-driven decision-making framework, is the on-site status of the potential intruder, and the weight coefficient , adjust according to actual conditions to ensure that the impact of each factor on the overall assessment is reasonable, and .

[0069] The development of this formula aims to solve the problem that the existing perimeter intrusion alarm system is not flexible and efficient enough in responding to complex and changeable potential intruders. Genetic algorithms and Bayesian networks are used for optimization and scenario simulation to ensure that the best alarm response combination can not only respond to intrusion incidents quickly and safely, but also reduce resource consumption and false alarm rate, and improve the overall performance and reliability of the system.

[0070] The following is a brief explanation of the design reasons for each item in the formula: The reason for designing the evaluation function is to comprehensively consider safety, timeliness, cost, dynamic response and false alarm rate, and to ensure that the alert response combination is both efficient and reliable by quantifying each factor and optimizing the weights.

[0071] The formula of the evaluation function is: :This sub-item emphasizes the safety of alert response measures by introducing weights and exponential factor , ensuring that safety is fully considered during the evaluation process, with higher and appropriate The value can improve the overall assessment score and encourage the selection of safer response measures; :This sub-item focuses on the speed of alert response, weight and exponential factor Ensure that the timeliness of responses is emphasized in the evaluation, with higher and appropriate The value can significantly improve the evaluation score and promote the selection of fast and effective response methods; :This item considers the cost-effectiveness of resource consumption and has a negative weight and exponential factor Emphasize the importance of reducing resource consumption, lower and appropriate The value helps to reduce the overall evaluation score, prompting the selection of response measures that optimize resource utilization; :This sub-item measures the dynamic adaptability of the response measures, with a weight of and exponential factor Ensure that the assessment focuses on the adaptability of response measures under different conditions, with higher and appropriate The value can significantly improve the assessment score and promote the selection of more flexible response measures; :This item aims to reduce the false positive rate, negative weight and exponential factor Emphasizing the importance of reducing false positives, lower and appropriate The value can significantly reduce the evaluation score, promoting the selection of response measures with lower false alarm rates; The purpose of adding up the sub-items is to comprehensively consider multiple evaluation dimensions and ensure that the final evaluation results can fully reflect the multi-faceted performance of the alert response combination. Each sub-item has a clear physical meaning and weight. By adding them together, a comprehensive score can be obtained to balance the relationship between different factors. : Evaluation result, indicating the combination of alert response A comprehensive evaluation score to measure its effectiveness; : Weight coefficients, which represent the weights of security score, timeliness score, cost score, dynamic response score and false alarm rate penalty. These weights are calculated to ensure that the sum is 1, and the importance of each factor is adjusted according to the actual situation; : Safety score, which evaluates the safety assurance capability of alert response measures. The higher the value, the safer it is. : Timeliness score, which measures the response speed of alert response measures. The higher the value, the more timely the response. : Cost score, which evaluates the resource consumption required for alert response measures. The higher the value, the lower the cost. : Dynamic response score, which measures the adaptability of the alert response measures to the behavior of potential intruders under different time and space conditions. The higher the value, the better the adaptability. : False alarm rate penalty, which measures the proportion of non-intrusion events that the system mistakenly identifies as intrusions. The higher the value, the lower the false alarm rate; : Exponential factor, which is used to adjust the sensitivity of each scoring item. It is calculated based on the frequency of security events in historical data, the dynamic change rate of the current environment, load level, visibility, false alarm rate and other parameters to adapt to different application scenarios.

[0072] The design reason of the calculation formula of the on-site status of the potential intruder is: design It aims to comprehensively evaluate the multi-dimensional behavioral characteristics of potential intruders, including movement paths, abnormal activity patterns, natural language processing results, and intelligence information. These factors are integrated through the weighted average method. It provides a comprehensive understanding of the on-site conditions, improves the accuracy and timeliness of the alert response, and ensures that the system can respond to potential threats quickly and effectively, while reducing the false alarm rate and enhancing the reliability and adaptability of the system.

[0073] The formula for calculating the on-site status of the potential intruder is expressed as: : Weight coefficients. These weights represent the relative importance of each factor in the comprehensive evaluation. They are adjusted according to the actual situation to ensure that the impact of each factor on the overall evaluation is reasonable. is the weight of the mobile path estimation result, is the weight of abnormal activity pattern detection, is the weight of the analysis result of natural language processing technology, is the weight of the information provided by the intelligence-driven decision-making framework.

[0074] The mobile path estimation result predicts the possible path of the potential intruder based on historical data and real-time monitoring. The score reflects the possible action routes that the potential intruder may take and their likelihood; : Abnormal activity pattern, detected abnormal activity pattern, used to trigger the alert response mechanism. This score measures the severity and urgency of the abnormal activity; : The result of natural language processing technology analysis; natural language processing technology is used to extract useful information from surveillance video or audio, such as identifying voice commands, abnormal sounds, etc. This score reflects the quality and relevance of the information obtained through natural language processing; : Information provided by the intelligence-driven decision-making framework, which provides background information about potential intruders and data to support decision-making through intelligence analysis. This score measures the quality and reliability of intelligence data.

[0075] The following is a brief explanation of how to obtain the parameters of the formula: Weight coefficient : Calculated based on historical data and dynamic changes in the current environment; Safety score : By analyzing the ability of alert response measures to contain potential intruders and the security effects on personnel and facilities, combined with the frequency of security events in historical data Based on the real-time monitoring information, a high score indicates good security, while a low score indicates the opposite, which can be obtained through expert evaluation, simulation testing and actual application feedback; : Weight coefficient, adjusted according to actual conditions to ensure that the impact of each factor on the overall evaluation is reasonable, is the weight of the mobile path estimation result, is the weight of abnormal activity pattern detection, is the weight of the analysis result of natural language processing technology, It is the weight of the information provided by the intelligence-driven decision-making framework, which can be obtained through expert evaluation, simulation testing and practical application feedback; Timeliness Rating : The response speed based on the alert response measures measures the time from when the system detects abnormal activity patterns to when effective countermeasures are taken. The higher the value, the more timely the response. It can be obtained through system logs, alarm records and user feedback statistics; Cost Rating : Evaluate the resource consumption required for alert response measures, including equipment cost, energy consumption, manpower input, etc. The higher the value, the lower the cost. This can be obtained through equipment operation logs, energy consumption records, and cost accounting; Dynamic Response Scoring : Measures the adaptability of alert response measures to the behavior of potential intruders under different time and space conditions. The higher the value, the better the adaptability. It can be obtained through scenario simulation, field testing and real-time monitoring data analysis; False Positive Rate Penalty : Measures the proportion of non-intrusion events that the system mistakenly identifies as intrusions. The higher the value, the lower the false alarm rate. The false alarm rate can be calculated through system logs, user feedback, and alarm records. Ecological event frequency in historical data: refers to the number of ecological environment-related events (such as animal invasions, plant damage, etc.) that have occurred in a specific period of time. The numerical value is obtained by extracting ecological event data from historical records or databases over a period of time and calculating its frequency; selecting a suitable time window (such as the past year, the past five years, etc.) for statistics; identifying long-term trends through time series analysis and predicting possible event frequencies in the future; The dynamic change rate of the current environment: indicates the speed at which various factors in the current environment (such as temperature, humidity, light intensity, etc.) change over time. The values ​​are monitored by the sensor network in real time and their change rates are calculated; the changes in environmental parameters in a short period of time are analyzed to identify the trend of rapid changes; the current change rate is compared with the long-term average value in historical data to evaluate the dynamics of the current environment; Load level: refers to the workload or pressure that the system bears when executing the expulsion measures, including resource consumption, equipment operation status, etc. The value evaluates the current load level by real-time monitoring of the system's energy consumption, equipment utilization rate and other indicators; regularly performs performance tests on the system and records the load conditions under different working conditions; collects user feedback on system performance and stability to understand the actual load conditions; False alarm rate: refers to the proportion of non-intrusion events that the system mistakenly identifies as intrusion events. The numerical value is obtained by analyzing system logs, counting the number and frequency of false alarm events; confirming which alarms are false alarms through field verification and user feedback; continuously improving the detection algorithm to reduce the false alarm rate, and recording the effect of each optimization; Visibility of potential intruders: refers to the possibility of potential intruders being clearly identified within the monitoring range, which is affected by factors such as lighting conditions and weather conditions. The value evaluates visibility through the quality of images captured by the monitoring camera, taking into account factors such as resolution, brightness, and contrast; analyzes the impact of lighting conditions and weather conditions (such as fog, rain, snow, etc.) on visibility; uses infrared cameras, thermal imagers and other technical means to improve visibility, and evaluates their effectiveness; Exponential Factor : Based on the frequency of security incidents in historical data , the dynamic change rate of the current environment , load level , visibility of potential intruders and false positive rate Calculated.

[0076] In the embodiment of the present application, in a large industrial park, the system predicts that humans are approaching the perimeter and showing abnormal activity patterns. Based on the estimated results of the moving path , the system immediately activates the multi-modal alert response mechanism. In order to select the best alert response combination ,The system uses the above formula to evaluate the effectiveness of each ,combination and optimizes the final solution through genetic algorithm and ,Bayesian network.

[0077] Parameter settings: Frequency of security incidents in historical data (H): 0.3 (the frequency of similar incidents in the past year); The dynamic change rate of the current environment (D): 0.6 (the current environment change rate); Load level (L): 0.7 (current system load level); Visibility of potential intruders (Visibility provided by cameras and sensors); False alarm rate (F): 0.1 (false alarm rate of the current system); Adjustment coefficients: ; The corresponding variable values ​​are: ; Calculate weight coefficient : ; Calculate the exponential factor : ; Calculate the evaluation function : Assumptions ; ; Evaluation Function The calculation result is 0.407, indicating that the overall effect of the current configuration of the alert response combination is good when considering security, timeliness, cost, dynamic response and false alarm rate. Specifically: Safety score A higher value means that the selected measures can effectively ensure safety; A higher timeliness score (T) indicates a fast response and the ability to respond to intrusions in a timely manner; The cost score (C) and false positive rate penalty (FPR) are low, reducing the possibility of resource waste and false positives; A high dynamic response score (DR) indicates that the measure can adapt to the behavioral changes of potential invaders under different temporal and spatial conditions.

[0078] Based on the evaluation results, the system selected the activation of sound and light alarms and drone patrols as the best warning response combination, and combined with Bayesian networks for scenario simulation to ensure the effectiveness and flexibility of the measures. This series of operations not only improved the sensitivity and reliability of the system, but also protected the ecological environment and achieved harmonious coexistence between man and nature.

[0079] Figure 2 A schematic diagram of the structure of a perimeter intrusion alarm system is provided for an embodiment of the present application, such as Figure 2 As shown, the device comprises: The recognition module 21 receives real-time environment interaction information from a plurality of distributed intelligent sensing nodes, recognizes image features in the real-time environment interaction information through a convolutional neural network, and analyzes time series behavior patterns in the real-time environment interaction information to obtain a behavior pattern recognition result of the real-time environment interaction information; An adjustment module 22 dynamically constructs a movement path of a potential intruder in a virtual three-dimensional geographic information system according to the behavior pattern recognition result, and adjusts the movement path prediction method in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; The analysis module 23, based on the movement path estimation result, immediately activates the multimodal alert response mechanism when an abnormal activity pattern is detected, analyzes the on-site status of the potential intruder using an intelligence-driven decision framework and natural language processing technology, and obtains the best alert response combination through a genetic algorithm; A generating module 24, based on the optimal warning response combination, generates harmless expulsion measures for the non-human biological intruders through an eco-friendly interference device when the non-human biological intruders invade, wherein the harmless expulsion measures are used to prevent the perimeter from being illegally crossed and to protect the ecological environment from being damaged; The optimization module 25, based on the actual effects of the optimal warning response combination and the harmless expulsion measures, uses the anonymized real-time environmental interaction information to learn group behavior patterns, and continuously optimizes the optimal warning response combination and the harmless expulsion measures to improve the sensitivity of the perimeter intrusion alarm.

[0080] Figure 2 The perimeter intrusion alarm system can be implemented Figure 1 The implementation principle and technical effect of the perimeter intrusion alarm method described in the embodiment are not described in detail. The specific way in which each module and unit performs operations in the perimeter intrusion alarm system in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0081] In one possible design, Figure 2 A perimeter intrusion alarm system of the illustrated embodiment may be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0082] The processing component 32 is used for the above Figure 1 A perimeter intrusion alarm method according to the embodiment.

[0083] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0084] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0085] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0086] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0087] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0088] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0089] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 A perimeter intrusion alarm method of the illustrated embodiment.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0091] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0092] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A perimeter intrusion alarm method, characterized in that: include: Receive real-time environment interaction information from multiple distributed intelligent perception nodes, identify image features in the real-time environment interaction information through a convolutional neural network, and parse the time series behavior pattern in the real-time environment interaction information to obtain a behavior pattern recognition result of the real-time environment interaction information; According to the behavior pattern recognition result, dynamically construct the movement path of the potential intruder in the virtual three-dimensional geographic information system, and adjust the prediction method of the movement path in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; Based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm; Based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure is generated by an eco-friendly interference device to prevent the perimeter from being illegally crossed and to protect the ecological environment from being damaged; Based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern is learned by using the anonymized real-time environmental interaction information, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to improve the sensitivity of the perimeter intrusion alarm.

2. The method according to claim 1, characterized in that: Based on the movement path estimation result, when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated, and the on-site status of the potential intruder is analyzed using an intelligence-driven decision framework and natural language processing technology, and the optimal alert response combination is obtained through a genetic algorithm, including: Based on the movement path estimation result, the movement path estimation result is analyzed by using anomaly detection and time series prediction, and when an abnormal activity pattern is detected, a multimodal alert response mechanism is immediately activated to generate an immediate response measure for the abnormal activity pattern; According to the immediate response measures, combining the intelligence-driven decision-making framework with natural language processing technology, using graph neural networks and knowledge graphs to analyze the on-site conditions of the potential intruder, so as to obtain an assessment report of the potential intruder; Based on the evaluation report of the potential intruder, the corresponding warning response combination is searched using genetic algorithm and particle swarm optimization, and the reinforcement learning strategy is used to continuously adjust and optimize according to the real-time feedback information of the actual situation of the potential intruder to generate the best warning response combination.

3. The method according to claim 2, characterized in that According to the immediate response measures, combined with the intelligence-driven decision-making framework and natural language processing technology, the on-site conditions of the potential intruder are analyzed using graph neural networks and knowledge graphs to obtain an assessment report on the potential intruder, including: Integrate and analyze the multi-source intelligence information collected in the immediate response measures on the potential intruder using an intelligence-driven decision-making framework according to the immediate response measures to generate a site situation analysis report including perimeter intrusion site environmental information and real-time monitoring information; Based on the on-site situation analysis report, using natural language processing technology to analyze the unstructured multi-source intelligence information in the on-site situation analysis report, and extracting the key information of the potential intruder to generate structured description information of the key information; Based on the structured description information, using a graph neural network, updating the knowledge graph of the potential intruder in the structured description information to obtain an updated knowledge graph; According to the entity relationships in the updated knowledge graph, the on-site conditions including the identity, capabilities and behavior trajectory of the potential intruder are identified and parsed to obtain an assessment report of the potential intruder.

4. The method according to claim 1, characterized in that According to the behavior pattern recognition result, the movement path of the potential intruder is dynamically constructed in the virtual three-dimensional geographic information system, and the prediction method of the movement path is adjusted in combination with the Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result, including: According to the behavior pattern recognition result, the behavior pattern recognition result is analyzed by using a virtual three-dimensional geographic information system to obtain a movement path model of each potential intruder; Based on the movement path model, using Bayesian optimization, evaluating and optimizing the movement path model, generating a preliminary movement path prediction method, so as to improve the accuracy of the movement path prediction of the potential intruder; Based on the preliminary moving path prediction method, a scenario simulation analysis process is performed on the reinforcement learning algorithm, and the preliminary moving path prediction method is further adjusted to obtain the moving path prediction method; Based on the prediction method of the moving path, the correction information of the prediction method fed back by reinforcement learning is used to analyze and correct the uncertainty and error in the moving path model to generate a moving path estimation result.

5. The method according to claim 4, characterized in that Based on the preliminary movement path prediction method, a scenario simulation analysis is performed on the reinforcement learning algorithm, and the preliminary movement path prediction method is further adjusted to obtain the movement path prediction method, including: Based on the preliminary movement path prediction method, the terrain conditions, weather conditions and emergency scenarios of the potential intruder are simulated to generate simulation scenarios including various movement paths; According to the simulation scenario, a reinforcement learning algorithm is applied to simulate the behavior pattern of the potential intruder, and the simulation result is compared and analyzed with the preset ideal behavior pattern of the potential intruder, the actual performance of the preliminary movement path prediction method is evaluated, and an analysis report of the preliminary movement path prediction method is obtained; Based on the analysis report, adjusting the reward mechanism and the penalty mechanism in the reinforcement learning algorithm to optimize the preliminary movement path prediction method and generate an improved movement path prediction method; Based on the improved mobile path prediction method, the improved mobile path prediction method is optimized through multiple rounds of iterative scenario simulation to obtain the mobile path prediction method that can adapt to various complex situations and dynamically respond to real-time environmental changes.

6. The method according to claim 1, characterized in that Based on the optimal warning response combination, when a non-human biological invader invades, a harmless expulsion measure for the non-human biological invader is generated through an eco-friendly interference device, and the harmless expulsion measure is used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged, including: Based on the trigger signal from the optimal alert response combination, using an intelligence-driven decision-making framework, analyzing and processing the real-time environmental interaction information of the trigger signal to obtain confirmation information that a non-human biological intruder has illegally crossed the perimeter; Based on the confirmed information, the selection and working mode of the eco-friendly interference device are evaluated, and the working mode with harmless stimulation that has been tested is selected to generate a setting plan for the repelling device; Using the repelling device setting scheme, the selected eco-friendly interference device is activated, and the repelling operation is performed according to the selected working mode to generate preliminary harmless repelling measures for the non-human biological intruder; Based on the preliminary harmless expulsion measures, feedback information collected from the real-time environmental interaction information of the non-human biological intruder is used to monitor the expulsion response of the non-human biological intruder to obtain a moving path tracking result of the non-human biological intruder; Based on the mobile path tracking results, the working mode of the eco-friendly interference device is dynamically adjusted to ensure that the eco-friendly interference device adapts to the behavioral characteristics of different types of non-human biological invaders to generate harmless expulsion measures, which are used to prevent the perimeter from being illegally crossed and protect the ecological environment from being damaged.

7. The method according to claim 1, characterized in that Based on the actual effects of the optimal warning response combination and the harmless expulsion measures, the group behavior pattern is learned by using the anonymized real-time environmental interaction information, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to improve the sensitivity of the perimeter intrusion alarm, including: Based on the optimal warning response combination and the actual effect of the harmless expulsion measures, the group behavior pattern of non-human biological intruders is studied by using the anonymized real-time environmental interaction information to obtain the behavioral characteristics of the non-human biological intruders; According to the behavior characteristics, the group behavior pattern of the non-human biological invaders is analyzed to generate a behavior model for describing the behavior characteristics; Based on the behavior model, combined with the intelligence-driven decision-making framework and the natural language processing framework, the response strategy of the optimal alert response combination is evaluated and analyzed to obtain an improvement plan for the response strategy; Based on the improvement plan of the response strategy, the identified improvement points are optimized, and the optimal warning response combination and the harmless expulsion measures are continuously optimized to ensure that the harmless expulsion measures are adaptable to different types of non-human biological intruders to improve the sensitivity of the perimeter intrusion alarm.

8. A perimeter intrusion alarm system, characterized in that: include: A recognition module receives real-time environment interaction information from a plurality of distributed intelligent sensing nodes, identifies image features in the real-time environment interaction information through a convolutional neural network, and analyzes time series behavior patterns in the real-time environment interaction information to obtain a behavior pattern recognition result of the real-time environment interaction information; An adjustment module dynamically constructs a movement path of a potential intruder in a virtual three-dimensional geographic information system according to the behavior pattern recognition result, and adjusts the movement path prediction method in combination with Bayesian optimization and reinforcement learning algorithms to generate a movement path estimation result; An analysis module, based on the movement path estimation result, immediately activates a multimodal alert response mechanism when an abnormal activity pattern is detected, analyzes the on-site status of the potential intruder using an intelligence-driven decision framework and natural language processing technology, and obtains an optimal alert response combination through a genetic algorithm; A generation module, based on the optimal warning response combination, generates harmless expulsion measures for the non-human biological intruders through an eco-friendly interference device when the non-human biological intruders invade, wherein the harmless expulsion measures are used to prevent the perimeter from being illegally crossed and to protect the ecological environment from being damaged; The optimization module uses the anonymized real-time environmental interaction information to learn group behavior patterns based on the actual effects of the optimal warning response combination and the harmless expulsion measures, and continuously optimizes the optimal warning response combination and the harmless expulsion measures to improve the sensitivity of the perimeter intrusion alarm.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a perimeter intrusion alarm method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a perimeter intrusion alarm method as described in any one of claims 1 to 7 is implemented.

Images