Smart park security system and control method thereof

Abstract

The invention discloses a smart park security and protection system and a control method thereof, and relates to the technical field of smart security and protection, the system comprises the following components: a diversified sensing module which integrates a video monitoring sensor, a face recognition sensor and a behavior analysis sensor and constructs an omnibearing sensing network. According to the technical scheme, all-dimensional and multi-angle monitoring and recognition of the park are achieved, the information acquisition capacity of the security and protection system is remarkably improved through the technological innovation, the comprehensiveness and accuracy of data are ensured, the diversified sensing module integrates multiple sensors of video monitoring, infrared detection, face recognition and behavior analysis, each sensor performs own functions and complements each other, and the monitoring and recognition efficiency is improved. According to the system, blind areas and errors possibly generated by a single sensor are effectively eliminated, the multi-angle sensing network enables the system to more accurately capture abnormal conditions in the park, and a solid foundation is provided for subsequent intelligent analysis and early warning.

Description

Technical Field

[0001] The present invention relates to the field of intelligent security technology, and in particular to a smart park security system and a control method thereof. Background Art

[0002] With the rapid development of science and technology, smart parks, as an important part of urban modernization construction, have increasingly prominent security management needs. As the core link to ensure park safety, smart park security systems are gradually developing in the direction of intelligence and integration. In this process, diversified perception and recognition technologies have become the key to improving the effectiveness of security systems. These technologies integrate multiple sensors and intelligent algorithms to achieve comprehensive monitoring and accurate identification of people, vehicles and environments in the park, providing strong technical support for the park's security management.

[0003] Traditional security systems have achieved monitoring and protection of industrial parks to a certain extent, but there are still obvious deficiencies in perception and recognition technology. First, most systems rely on a single video surveillance sensor, which makes it difficult to ensure the comprehensiveness and accuracy of information in complex environments, resulting in monitoring blind spots and errors. Secondly, traditional systems mainly rely on manual judgment to identify abnormal behaviors, which is not only inefficient, but also easily affected by human factors, prone to false alarms and missed reports. In addition, for multi-dimensional and multi-level perception data, traditional systems lack effective integration and analysis mechanisms, making it difficult to fully tap the value of data and provide strong support for security management.

[0004] Therefore, a smart campus security system and its control method are developed to solve the shortcomings of traditional security systems through technological innovation and provide more comprehensive, efficient and intelligent technical support for the security management of the campus. Summary of the invention

[0005] The purpose of this invention is to make up for the shortcomings of the existing technology and provide a smart park security system and its control method, which can realize all-round and multi-angle monitoring and identification of people, vehicles and environment in the park by integrating multiple sensor technologies and AI algorithms. This technology not only improves the information acquisition capability and identification accuracy of the security system, but also enhances the intelligence level and early warning capability of the system, providing a more reliable and efficient technical guarantee for the security management of the park.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: On the one hand, a smart park security system includes the following components:

[0007] Diversified perception module: Integrate sensors for video surveillance, face recognition, and behavior analysis to build a comprehensive perception network. Through advanced data fusion algorithms, data from different sensors are integrated to improve the comprehensiveness and accuracy of information.

[0008] Intelligent recognition module: Based on the collected video data, it uses deep learning algorithms for high-precision face recognition, quickly and accurately identifies people in the park, implements real-time warnings for blacklisted people, and automatically identifies and marks abnormal behaviors through behavioral analysis technology, improving the early warning capabilities of security incidents;

[0009] Abnormal detection and early warning module: Combined with the intelligent recognition module, it conducts in-depth analysis of images and data in the surveillance video, automatically identifies and marks potential security threats, and sets up a multi-level early warning mechanism according to the severity of abnormal behavior;

[0010] Decision support module: continuously collects and stores various security data, builds a rich historical database, and uses big data analysis technology to conduct in-depth mining of historical data to discover potential safety hazards and patterns;

[0011] Remote monitoring and emergency command module: supports remote viewing of real-time monitoring images of the park through terminal devices, enabling monitoring of the park at any time. When a security incident occurs, it automatically starts the emergency command process, quickly dispatches various resources, and conducts effective response and disposal.

[0012] Furthermore, the diversified perception module integrates perception devices such as video surveillance cameras, face recognition sensors, and behavior analysis sensors, and is installed in key areas of the park to form a broad and diverse perception network. Sensors are used to collect data on video images, facial features, and behavior patterns in real time, and the original data is denoised, enhanced, and formatted in a unified manner. A Gaussian mixture model is used to integrate data from different sensors to eliminate redundancy and conflict and retain unique information.

[0013] Furthermore, the diversified perception module uses a Gaussian mixture model to integrate data from different sensors and represents the data from different sensors as a set of observations {x1, x2, ..., x n}, where each observation x i is a d-dimensional vector, and the Gaussian mixture model is expressed as: Where K is the number of Gaussian distributions, π k is the mixing coefficient of the kth Gaussian distribution, satisfying N(x|μ k ,Σ k ) is the kth Gaussian distribution, μ k is the mean, Σ k is the covariance matrix, initialize the mixing coefficient π k , mean μ k and the covariance matrix Σ k Parameters, calculate each observation x i The posterior probability generated by each Gaussian component is: Update the model parameters, the formula is: Mixing coefficient: Mean: Covariance matrix: Repeat the steps of updating the posterior probability and model parameters until the model converges, and use the Gaussian mixture model to integrate data from different sensors, eliminate redundancy and conflict, and generate a more comprehensive information description.

[0014] Furthermore, the intelligent recognition module builds a high-precision face recognition model based on the training of a large amount of face image data based on the convolutional neural network, and applies the model in real time to perform face detection in the video stream data within the park, extracts features and compares them with the personnel database to identify the identity, and compares them with the blacklist database at the same time. If a blacklisted person is found, an early warning is immediately triggered, and a behavior analysis model is built based on the support vector machine algorithm using the normal and abnormal behavior pattern data sets to analyze the behavior of the people in the surveillance video, the trajectory, speed, and posture of the behavior, and automatically identify and mark abnormal behavior.

[0015] Furthermore, the intelligent recognition module builds a high-precision face recognition model based on a convolutional neural network. Assume that the collected face images are {(x1, y1), (x2, y2), ..., (x N ,y N )}, where x i is the i-th input face image, y i is the corresponding label, N means there are N samples, the convolutional neural network is composed of convolutional layer, pooling layer and fully connected layer, initialize the weight W and bias b of the convolutional neural network, for the input face image x i After a series of convolution and pooling operations, and the final fully connected layer, the output prediction value is obtained The convolution operation is to extract features from the input image. The calculation formula is: in, is the neuron output of the i-th row and j-th column of the l-th layer, f is the activation function, is the weight, P and Q are the sizes of the convolution kernel, is a bias, which is used to reduce the spatial size of the feature map through maximum pooling while retaining the main features. The maximum pooling operation formula is: Among them, R ijIt is the area corresponding to the pooling window. After multiple convolution and pooling layers, the feature map obtained will be flattened and input into the fully connected layer for further processing and integration of the extracted features. The calculation formula of the fully connected layer is: Z = WX + b, where W is the weight matrix and b is the bias vector. The stochastic gradient descent algorithm is used to minimize the loss function, update the parameters of the model, use real-time data to calculate the accuracy evaluation results, adjust the number of convolution kernels, the number of layers, and learning of the model, and perform multiple iterative training to obtain the optimal model performance.

[0016] Furthermore, the intelligent recognition module constructs a behavior analysis model based on the normal and abnormal behavior pattern data sets through the support vector machine algorithm. The support vector machine is to find an optimal hyperplane in the feature space to separate the normal and abnormal behavior data. The hyperplane formula is calculated as: T x+b=0, where w is the normal vector of the hyperplane, b is the intercept, and x is the eigenvector of the input data. The optimization goal of SVM is to maximize the interval and find w and b that meet the following conditions. The formula is: Among them, y i is the class label of the sample, normal behavior is usually marked as 1, abnormal behavior is marked as -1, x i is the feature vector of the i-th sample, n is the number of samples, and the optimal w and b are calculated. The model is trained using the training set, and the trained SVM model is evaluated using real-time data. According to the evaluation results, the parameters σ and penalty coefficient C of the model kernel function are adjusted, and the adjusted SVM behavior analysis model is applied to actual scenarios. New data is monitored in real time to determine whether it is normal behavior or abnormal behavior.

[0017] Furthermore, the anomaly detection and early warning module conducts in-depth analysis of the park's real-time data through the constructed face recognition model and behavior analysis model, automatically identifies potential security threats, evaluates the detected anomalies, determines their severity and type, and sets low, medium, and high-level early warning mechanisms according to the severity and type of abnormal behavior. When the early warning is triggered, corresponding measures are taken and feedback information is collected, and continuous optimization is carried out based on application effects and feedback.

[0018] Furthermore, the decision support module continuously collects relevant data from various security equipment and information systems, formats the data and then integrates and stores them in a historical database, performs data preprocessing and storage, and builds a big data analysis platform. It identifies hidden dangers, patterns and trends, gives relevant suggestions based on business knowledge and expert experience, organizes the analysis results and suggestions into easy-to-understand reports, and uses data visualization tools to display the analysis results.

[0019] Furthermore, the remote monitoring and emergency command module allows users to remotely access the monitoring system through terminal devices and view the images of various monitoring points in the park in real time. The system can automatically identify and analyze the monitoring images through the built-in algorithm model, detect abnormal behavior, automatically mark and highlight the abnormal area, and generate an alarm push. After receiving the alarm, the system automatically confirms the authenticity and severity of the incident, and activates the corresponding emergency plan based on the event type and evaluation results. It automatically calculates and plans the optimal resource scheduling plan, and establishes a real-time communication link between the site and the command center through voice communication, tracks the event handling process, records key information, summarizes and analyzes the incident afterwards, evaluates the effectiveness and shortcomings of the emergency plan, and provides a basis for future optimization.

[0020] On the other hand, a control method of a smart park security system is provided, and the specific steps of the control method are as follows:

[0021] Diversified perception data collection: Video surveillance, infrared detection, face recognition, and behavior analysis sensors are deployed in the park to form a multi-level perception network. The sensors collect data on people, vehicles, and the environment in the park in real time and transmit it to the system's processing center.

[0022] Data preprocessing and fusion: Preprocess the collected raw data and use data fusion technology to integrate data from different sensors to form a unified view of the park's security situation;

[0023] AI intelligent analysis: AI algorithms are introduced to perform intelligent analysis on pre-processed data. Face recognition algorithms are used to identify the identities of people entering and leaving the park and compare them with the blacklist database. Behavior analysis algorithms are used to automatically detect abnormal behaviors in surveillance videos.

[0024] Early warning and decision support: Based on the results of AI analysis, early warning information is generated in real time and notified to park managers through sound and light alarms and SMS notifications. Decision support functions are provided to provide managers with scientific response strategies based on the early warning level and the actual situation of the park.

[0025] Linked response and closed-loop management: Once an early warning is triggered, the security equipment will be automatically linked and appropriate emergency measures will be taken to form a closed-loop management mechanism to record and track the event handling process.

[0026] Compared with the prior art, the smart park security system and control method thereof have the following beneficial effects:

[0027] 1. The present invention realizes all-round and multi-angle monitoring and identification of the park by introducing diversified perception and recognition technologies. This technological innovation significantly improves the information acquisition capability of the security system and ensures the comprehensiveness and accuracy of the data. The diversified perception module integrates multiple sensors such as video surveillance, infrared detection, face recognition, and behavior analysis. Each sensor performs its own functions and complements each other, effectively eliminating the blind spots and errors that may be caused by a single sensor. This multi-angle perception network enables the system to capture abnormal situations in the park more accurately, providing a solid foundation for subsequent intelligent analysis and early warning.

[0028] 2. The diversified perception and recognition technology of the present invention not only improves the recognition ability of the security system, but also enhances the intelligence level of the system. By combining AI algorithms, the system can intelligently analyze abnormal behaviors in surveillance videos, and automatically identify and mark abnormal behaviors in surveillance videos through behavioral analysis technology. This intelligent analysis capability enables the system to respond to security incidents more quickly and provide timely warning information for park managers. The diversified perception data also provides a rich data source for the decision support module of the system, enabling the system to explore potential safety hazards and patterns based on big data analysis technology, and provide a more scientific decision-making basis for the safety management of the park.

[0029] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 It is a process operation diagram of a smart park security system;

[0032] Figure 2 The present invention is a flow chart of a control method for a smart park security system. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Embodiment 1:

[0035] This embodiment provides a smart campus security system, which builds an all-round, multi-angle security monitoring network by introducing diversified perception and recognition technologies, including video surveillance, infrared detection, face recognition and behavior analysis. It not only significantly improves the comprehensiveness and accuracy of security information, but also combines AI algorithms to realize intelligent analysis and rapid warning of abnormal behaviors, effectively reducing false alarm and missed alarm rates. In addition, the rich perception data provides a solid data foundation for the decision support module, enabling the system to explore potential safety hazards based on big data analysis and provide scientific decision-making basis for campus safety management.

[0036] The diversified perception module integrates the perception devices of video surveillance cameras, face recognition sensors, and behavior analysis sensors, and is deployed in key areas of the park to form a wide and diverse perception network. The sensors are used to collect data on video images, face features, and behavior patterns in real time, and the data from different sensors are represented as a set of observation values ​​{x1, x2, ..., x n}, where each observation x i is a d-dimensional vector, and the Gaussian mixture model is used for data fusion. The formula of the Gaussian mixture model is: Among them, K is the number of Gaussian distributions, π k is the mixing coefficient of the kth Gaussian distribution, satisfying N(x|μ k ,Σ k ) is the kth Gaussian distribution with mean μ k , the covariance matrix is ​​Σ k , initialize the mixing coefficient π k , mean μ k and the covariance matrix Σ k Parameters, calculated for each observation x i The posterior probability generated by each Gaussian component is: Update model parameters: Mixing coefficient: Mean: Covariance matrix: The steps of obtaining posterior probabilities and updating model parameters are repeated until the model converges, and the data from different sensors are fused together.

[0037] The intelligent recognition module uses the camera to capture the video stream in real time, trains it through the convolutional neural network, builds a high-precision face recognition model, and deploys the trained model in the data center. It performs real-time face detection on the video stream captured by the camera, extracts the detected facial features, compares them with the park personnel database, and quickly identifies the identity. At the same time, it compares it with the blacklist database and immediately triggers an alert when a blacklisted person is found. It collects and annotates data sets of normal and abnormal behavior patterns, and uses the support vector machine (SVM) algorithm to train a behavior analysis model based on the annotated data set. It continuously analyzes the behavior of people in the video, automatically identifies and marks abnormal behavior by tracking their trajectory, calculating movement speed and posture changes. Once blacklisted people and abnormal behavior are identified, the system automatically triggers an alert and automatically starts the emergency response mechanism according to the alert level.

[0038] The anomaly detection and early warning module uses the created face recognition model and behavior analysis model to conduct multi-dimensional analysis of the detected behavior in terms of trajectory, speed, and posture, identify abnormal behavior, and evaluate the severity and type of abnormal behavior. Different abnormal plans are initiated for different abnormal behaviors. For minor abnormal behaviors, such as slow wandering, the relevant personnel are reminded through park broadcasts and APP push notifications. For moderate abnormal behaviors, such as fast running and gathering, security personnel are automatically notified to go to the scene to check and video surveillance of relevant areas is turned on. For serious abnormal behaviors, such as fighting and intrusion, sound and light alarms are immediately triggered. When abnormal behavior is found, an emergency alarm signal is sent to the park management center and the emergency plan is activated. According to the warning level, corresponding emergency response measures are taken, and statistical analysis is performed on the warning records and feedback information to identify system performance bottlenecks and reasons for false alarms and missed alarms. According to the analysis results, the parameters of the behavior detection model are adjusted to improve recognition accuracy and efficiency.

[0039] The decision support module continuously collects relevant data from various security equipment and information systems, formats the collected data, and integrates and stores them in a historical database. It performs preliminary processing on the integrated data and encrypts sensitive data to ensure the security of data transmission and storage. It builds a cloud computing-based big data analysis platform to achieve rapid processing and efficient analysis of massive security data. For the identified safety hazards, behavior patterns, and development trends, the analysis results are organized into easy-to-understand reports, and the analysis results are displayed in the form of charts using data visualization tools, so that management and security personnel can quickly grasp the safety status of the park.

[0040] The remote monitoring and emergency command module integrates all surveillance cameras and sensors in the park to ensure that all important areas can be effectively monitored. Users can remotely access the monitoring system through terminal devices and view the real-time images of each monitoring point in the park at any time. The system's built-in intelligent analysis model analyzes the monitoring images in real time and can accurately identify abnormal behaviors. Once abnormal behaviors are detected, the abnormal areas are automatically marked and highlighted in the monitoring images to attract the user's attention. Alarm information is sent to relevant personnel through preset alarm channels, and the authenticity and severity of the confirmed events are evaluated. According to the event type and evaluation results, the corresponding emergency plan is automatically started, and the optimal resource scheduling plan is planned. Through voice communication, a real-time communication link between the site and the command center can be quickly established. The command center personnel track the event handling process in real time, obtain first-hand information on the site, and guide the on-site personnel to take correct response measures. After the event is handled, an event summary report is automatically generated to evaluate the effectiveness and shortcomings of this emergency plan, providing an important basis for future plan optimization.

[0041] Embodiment 2:

[0042] Based on the first embodiment of this embodiment, the application of diversified perception and recognition technologies is particularly emphasized:

[0043] Video surveillance cameras, facial recognition sensors, and behavioral analysis sensors are deployed in key areas of the park to form a broad and diverse perception network, ensuring that all equipment can work normally and transmit data to the system control center in real time. Sensors are used to collect video images, facial features, and behavioral pattern data in real time, and the raw data is denoised, enhanced, and formatted in a unified manner to improve the accuracy and efficiency of subsequent analysis.

[0044] The data from different sensors are represented as a set of observations {x1,x2,...,x n}, where each observation x i is a d-dimensional vector, and the Gaussian mixture model is used to integrate different sensor data. The Gaussian mixture model is expressed as: Among them, K is the number of Gaussian distributions, π k is the mixing coefficient of the kth Gaussian distribution, satisfying is the kth Gaussian distribution, μ k is the mean, Σ k is the covariance matrix, initialize the mixing coefficient π k , mean μ k and the covariance matrix Σ k Parameters, calculated for each observation x i The posterior probability generated by each Gaussian component is: Update the model parameters, the formula is: Mixing coefficient: Mean: Covariance matrix: Repeat the above steps until the model converges, and obtain a GMM model that can effectively integrate different sensor data, eliminate redundancy and conflict, and generate a more comprehensive and accurate information description.

[0045] A high-precision face recognition model is built based on a convolutional neural network. A large amount of face image data is collected as training data, and the corresponding identity labels are annotated. A convolutional neural network (CNN) model is built, including a convolutional layer, a pooling layer, and a fully connected layer. The weight W and bias b of the convolutional neural network are initialized. For the input face image x i After a series of convolution and pooling operations, and the final fully connected layer, the output prediction value is obtained The convolution operation is to extract features from the input image. The calculation formula is: in, is the neuron output of the i-th row and j-th column of the l-th layer, f is the activation function, is the weight, P and Q are the sizes of the convolution kernel, is a bias, which is used to reduce the spatial size of the feature map through maximum pooling while retaining the main features. The maximum pooling operation calculation formula is: Among them, R ij It is the area corresponding to the pooling window. After multiple convolution and pooling layers, the obtained feature map will be flattened and input into the fully connected layer for further processing and integration of the extracted features. The calculation formula of the fully connected layer is: Z=WX+b, where W is the weight matrix and b is the bias vector. The random gradient descent algorithm is used to optimize the model parameters. The optimal model performance is obtained through multiple iterative training. The model is applied in real time to perform face detection and recognition in the video stream captured by the cameras deployed in the park. The detected facial features are compared with the personnel database to identify the identity, and compared with the blacklist database to trigger an early warning.

[0046] Collect normal and abnormal behavior pattern data sets, and use the support vector machine (SVM) algorithm to build a behavior analysis model. The support vector machine finds the optimal hyperplane in the feature space to distinguish normal and abnormal behaviors. The hyperplane formula is calculated as: T x+b=0, where w is the normal vector of the hyperplane, b is the intercept, and x is the eigenvector of the input data. The optimization goal of SVM is to maximize the interval and find w and b that meet the following conditions. The formula is: Among them, y i is the class label of the sample, normal behavior is usually marked as 1, abnormal behavior is marked as -1, x iis the feature vector of the i-th sample, n is the number of samples, and the optimal w and b are calculated. The model is evaluated using the training set, and the model parameters are adjusted according to the evaluation results to obtain better performance. The adjusted SVM behavior analysis model is applied to actual scenarios to monitor the behavior of people in surveillance videos in real time, analyze the characteristics of the trajectory, speed, and posture of the behavior, and automatically identify and mark abnormal behavior.

[0047] The diversified perception module and the intelligent recognition module are seamlessly integrated to form a complete smart campus security system. The system receives and processes data from various perception devices in real time, generates more comprehensive and accurate information descriptions through data fusion algorithms, and uses face recognition models and behavior analysis models to intelligently identify and analyze surveillance videos, timely discover and warn of potential security threats, support real-time communication and event tracking functions, ensure that the command center can respond to and handle various security incidents quickly, summarize and analyze incidents afterwards, evaluate the effectiveness and shortcomings of emergency plans, and provide a basis for future optimization.

[0048] Embodiment three:

[0049] Based on the first and second embodiments, the specific process steps of the system are as follows:

[0050] (1) Deploy a variety of sensing devices to form a wide sensing network, collect and pre-process raw data, and use Gaussian mixture models to integrate data to ensure the comprehensiveness and accuracy of information;

[0051] (2) By building a high-precision face recognition and behavior analysis model, we can quickly identify the identities of people in the park and automatically detect abnormal behaviors, thereby improving the intelligence level of the security system;

[0052] (3) Use behavioral analysis models to deeply analyze monitoring data and set up a multi-level early warning mechanism to promptly detect and respond to potential security threats to ensure the safety and stability of the park;

[0053] (4) Collect and integrate security data, use big data analysis technology to explore potential safety hazards, provide managers with scientific decision-making support, and improve the efficiency of park safety management;

[0054] (5) Realize remote monitoring functions, automatically identify abnormal behaviors through intelligent analysis models, and establish real-time communication links to ensure rapid response and effective disposal in emergency situations;

[0055] (6) Conduct comprehensive testing of the system to ensure the stability and reliability of system functions, and make optimization adjustments based on test results and user feedback to improve user experience;

[0056] (7) Ensure the daily stable operation of the system, perform regular system maintenance, provide user training support, and ensure the long-term effective operation of the security system.

[0057] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A smart park security system, characterized in that: The system consists of the following components: Diversified perception module: Integrate sensors for video surveillance, face recognition, and behavior analysis to build a comprehensive perception network. Through advanced data fusion algorithms, data from different sensors are integrated to improve the comprehensiveness and accuracy of information. Intelligent recognition module: Based on the collected video data, it uses deep learning algorithms to perform high-precision face recognition, quickly and accurately identify people in the park, implement real-time warnings for blacklisted people, and automatically identify and mark abnormal behaviors through behavioral analysis technology to improve the early warning capabilities of security incidents; Abnormal detection and early warning module: Combined with the intelligent recognition module, it conducts in-depth analysis of images and data in the surveillance video, automatically identifies and marks potential security threats, and sets up a multi-level early warning mechanism according to the severity of abnormal behavior; Decision support module: continuously collects and stores various security data, builds a rich historical database, and uses big data analysis technology to conduct in-depth mining of historical data to discover potential safety hazards and patterns; Remote monitoring and emergency command module: supports remote viewing of real-time monitoring images of the park through terminal devices, enabling monitoring of the park at any time. When a security incident occurs, it automatically starts the emergency command process, quickly dispatches various resources, and conducts effective response and disposal.

2. A smart park security system according to claim 1, characterized in that: The diversified perception module integrates the perception devices of video surveillance cameras, face recognition sensors, and behavior analysis sensors, and is installed in key areas of the park to form a broad and diverse perception network. It uses sensors to collect data on video images, facial features, and behavior patterns in real time, denoises, enhances, and unifies the format of the original data, and uses a Gaussian mixture model to integrate data from different sensors, eliminate redundancy and conflict, and retain unique information.

3. A smart park security system according to claim 2, characterized in that: The diversified perception module uses a Gaussian mixture model to integrate data from different sensors and represents the data from different sensors as a set of observations {x1, x2, ..., x n }, where each observation x i is a d-dimensional vector, and the Gaussian mixture model is expressed as: Among them, K is the number of Gaussian distributions, π k is the mixing coefficient of the kth Gaussian distribution, satisfying N(x|μ k ,Σ k ) is the kth Gaussian distribution, μ k is the mean, Σ k is the covariance matrix, initialize the mixing coefficient π k , mean μ k and the covariance matrix Σ k Parameters, calculated for each observation x i The posterior probability generated by each Gaussian component is: Update the model parameters, the formula is: Mixing coefficient: Mean: Covariance matrix: Repeat the steps of updating the posterior probability and model parameters until the model converges, and use the Gaussian mixture model to integrate data from different sensors, eliminate redundancy and conflict, and generate a more comprehensive information description.

4. The smart park security system according to claim 1, characterized in that: The intelligent recognition module builds a high-precision face recognition model based on the training of a large amount of face image data by convolutional neural network, applies the model in real time to face detection in the video stream data in the park, extracts features and compares them with the personnel database to identify the identity, and compares them with the blacklist database at the same time. If a blacklisted person is found, an early warning will be triggered immediately. The behavior analysis model is built based on the support vector machine algorithm using the normal and abnormal behavior pattern data set to analyze the trajectory, speed, and posture of the personnel in the monitoring video, and automatically identify and mark abnormal behavior.

5. A smart park security system according to claim 4, characterized in that: The intelligent recognition module builds a high-precision face recognition model based on a convolutional neural network. Suppose the collected face images are {(x1, y1), (x2, y2), ..., (x N ,y N )}, where x i is the i-th input face image, y i is the corresponding label, N means there are N samples, the convolutional neural network is composed of convolutional layer, pooling layer and fully connected layer, initialize the weight W and bias b of the convolutional neural network, for the input face image x i After a series of convolution and pooling operations, and the final fully connected layer, the output prediction value is obtained The convolution operation is to extract features from the input image. The calculation formula is: in, is the neuron output of the i-th row and j-th column of the l-th layer, f is the activation function, is the weight, P and Q are the sizes of the convolution kernel, is a bias, which is used to reduce the spatial size of the feature map through maximum pooling while retaining the main features. The maximum pooling operation formula is: Among them, R ij It is the area corresponding to the pooling window. After multiple convolution and pooling layers, the feature map obtained will be flattened and input into the fully connected layer for further processing and integration of the extracted features. The calculation formula of the fully connected layer is: Z = WX + b, where W is the weight matrix and b is the bias vector. The stochastic gradient descent algorithm is used to minimize the loss function, update the parameters of the model, use real-time data to calculate the accuracy evaluation results, adjust the number of convolution kernels, the number of layers, and learning of the model, and perform multiple iterative training to obtain the optimal model performance.

6. The smart park security system according to claim 4, characterized in that: The intelligent recognition module constructs a behavior analysis model based on the normal and abnormal behavior pattern data sets through the support vector machine algorithm. The support vector machine is to find an optimal hyperplane in the feature space to separate the normal and abnormal behavior data. The hyperplane formula is calculated as: T x+b=0, where w is the normal vector of the hyperplane, b is the intercept, and x is the eigenvector of the input data. The optimization goal of SVM is to maximize the interval and find w and b that meet the following conditions. The formula is: Among them, y i is the class label of the sample, normal behavior is usually marked as 1, abnormal behavior is marked as -1, x i is the feature vector of the i-th sample, n is the number of samples, and the optimal w and b are calculated. The model is trained using the training set, and the trained SVM model is evaluated using real-time data. According to the evaluation results, the parameters σ and penalty coefficient C of the model kernel function are adjusted, and the adjusted SVM behavior analysis model is applied to actual scenarios. New data is monitored in real time to determine whether it is normal behavior or abnormal behavior.

7. The smart park security system according to claim 1, characterized in that: The anomaly detection and early warning module conducts in-depth analysis of the park's real-time data through the constructed face recognition model and behavior analysis model, automatically identifies potential security threats, evaluates the detected anomalies, determines their severity and type, and sets low, medium, and high-level early warning mechanisms according to the severity and type of abnormal behavior. When the early warning is triggered, corresponding measures are taken and feedback information is collected, and continuous optimization is carried out based on application effects and feedback.

8. The smart park security system according to claim 1, characterized in that: The decision support module continuously collects relevant data from various security equipment and information systems, formats the data and integrates and stores them in a historical database, performs data preprocessing and storage, and builds a big data analysis platform. It identifies hidden dangers, patterns and trends, combines business knowledge and expert experience to give relevant suggestions, organizes the analysis results and suggestions into easy-to-understand reports, and uses data visualization tools to display the analysis results.

9. The smart park security system according to claim 1, characterized in that: The remote monitoring and emergency command module allows users to remotely access the monitoring system through terminal devices and view the images of various monitoring points in the park in real time. The system can automatically identify and analyze the monitoring images through the built-in algorithm model, detect abnormal behavior, automatically mark and highlight the abnormal area, and generate an alarm push. After receiving the alarm, the system automatically confirms the authenticity and severity of the incident, and activates the corresponding emergency plan based on the event type and evaluation results. It automatically calculates and plans the optimal resource scheduling plan, and establishes a real-time communication link between the site and the command center through voice communication, tracks the event handling process, records key information, summarizes and analyzes the incident afterwards, evaluates the effectiveness and shortcomings of the emergency plan, and provides a basis for future optimization.

10. A control method for a smart park security system, characterized in that: The specific steps of the control method are: Diversified perception data collection: Video surveillance, infrared detection, face recognition, and behavior analysis sensors are deployed in the park to form a multi-level perception network. The sensors collect data on people, vehicles, and the environment in the park in real time and transmit it to the system's processing center. Data preprocessing and fusion: Preprocess the collected raw data and use data fusion technology to integrate data from different sensors to form a unified view of the park's security situation; AI intelligent analysis: AI algorithms are introduced to perform intelligent analysis on pre-processed data. Face recognition algorithms are used to identify the identities of people entering and leaving the park and compare them with the blacklist database. Behavior analysis algorithms are used to automatically detect abnormal behaviors in surveillance videos. Early warning and decision support: Based on the results of AI analysis, early warning information is generated in real time and notified to park managers through sound and light alarms and SMS notifications. Decision support functions are provided to provide managers with scientific response strategies based on the early warning level and the actual situation of the park. Linked response and closed-loop management: Once an early warning is triggered, the security equipment will be automatically linked and appropriate emergency measures will be taken to form a closed-loop management mechanism to record and track the event handling process.

Images