Intelligent security information management method and device, and storage medium

Abstract

The application provides a kind of intelligent security information management method, device and storage medium, it is related to intelligent security technical field, comprising: collecting and detecting the clear image data of community key area, extracting face area and cutting storage as independent image, using pre-trained deep learning model to the feature extraction of extracted face, and with registered database is compared, by calculating the Euclidean distance between feature vectors to determine personnel type, after being identified as registered personnel, further judge whether it is temporary personnel, and monitor the residence time;If exceed the set threshold, then send out alarm, while tracking the location of unregistered personnel.Real-time analysis of video stream, monitor the behavior of registered personnel, if abnormal behavior is found, automatically alarm and record relevant image, build self-learning mechanism, update database in time through the feedback of security personnel, the scheme provides a more intelligent solution for community safety management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent security and protection, in particular to an intelligent security and protection information management method and device and a storage medium. BACKGROUND

[0002] In modern society, security and protection monitoring systems have become an important part of maintaining public safety and are widely used in communities, shopping malls, transportation hubs, and other places. Traditional security and protection monitoring systems usually rely on fixed cameras to record videos and rely on manual patrols by security personnel to identify potential safety hazards. Although this method can achieve monitoring of abnormal events to some extent, it has obvious shortcomings. First, manual patrols are not only inefficient but also limited by the experience and judgment of security personnel, leading to frequent false negatives and false positives. Second, traditional monitoring systems often struggle to accurately identify dynamic scenes and complex behaviors, especially in situations with high personnel flow or poor lighting conditions, where image quality and recognition accuracy are affected. In addition, traditional systems lack self-learning capabilities and cannot dynamically adjust to changing environments and user needs, resulting in lagging technology updates and difficulty in meeting growing security needs.

[0003] On the technical level, although some monitoring systems have begun to incorporate basic image processing techniques and simple behavior analysis algorithms, these technologies still have limitations. These algorithms often rely on fixed rules and thresholds and cannot effectively handle diverse behavior patterns, resulting in difficulties in adapting to complex security scenarios in practical applications. In addition, existing feature extraction and comparison methods are inefficient and lack real-time performance when dealing with high-dimensional feature spaces. Once the system encounters new behavior patterns or abnormal events, it often lacks the ability to respond quickly and cannot provide timely warnings. In summary, traditional security and protection monitoring technology has shortcomings in terms of intelligence, accuracy, real-time performance, and adaptability, and more advanced technical means are needed to improve the overall effectiveness and reliability of security and protection management.

[0004] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide an intelligent security and protection information management method and system to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides the following technical solutions:

[0007] The intelligent security and protection information management method includes the following specific steps:

[0008] Step 1: Collect clear image data of key areas in the community, perform face detection on the collected images, use deep neural networks to identify and extract face regions, obtain the coordinates of the face by applying detection algorithms to each frame of the image, and crop and store it as an independent face image;

[0009] Step 2: Use a pre-trained deep learning model to extract features from the extracted face images and compare them with the stored registered database. By calculating the Euclidean distance between the extracted feature vector and the registered vector, determine the type of person and classify them into two categories: registered and unregistered.

[0010] Step 3: Classify the image information according to the recognition results. If it is identified as a registered person, further determine whether it is a temporary person and calculate its stay time. If it exceeds the set threshold, an alarm will be issued. At the same time, obtain the location information of unregistered persons and track them.

[0011] Step 4: Analyze the video stream in real time, monitor the inappropriate behavior of registered personnel, use behavior recognition algorithms to predict behavior, and if abnormal behavior is detected, the system will automatically issue an alarm and record relevant image information for subsequent analysis.

[0012] Step 5: Build a self-learning mechanism. Security personnel provide feedback on each alarm, update the database in a timely manner, and regularly retrain the model with new data to improve recognition accuracy and system adaptability.

[0013] Furthermore, the key areas include the community gates, garage entrances, and the community's parks, squares, and children's playgrounds;

[0014] Images are acquired from a camera or video stream and preprocessed, including resizing and normalization. Face images are extracted from the acquired image information and face detection is performed using the Caffe model provided by OpenCV.

[0015] For each frame of image The expression for facial region localization is as follows:

[0016]

[0017] in, For the current frame image, The coordinates of the detected face region, specifically the coordinates of the top left and bottom right corners. This is a function call that represents a face detection operation performed on the current frame image;

[0018] Based on the detected Coordinates, draw a rectangular frame on the original map to identify the face, according to these coordinates from the original image to cut out the face area, and store it as an independent face image, denoted as .

[0019] Further, using a pre-trained deep learning model for face feature extraction, the expression is as follows:

[0020]

[0021] Wherein, is the extracted face feature vector, containing various attributes of the face, is the cropped face image, indicates that the pre-trained deep learning model is called to forward propagate to extract face features; The various attributes of the face are: facial contour, eye position, nose shape, mouth feature and skin texture feature;

[0022] According to the following formula, calculate the Euclidean distance between the extracted feature vector and the registered feature vector:

[0023]

[0024] Wherein, is the Euclidean distance value, indicating the similarity of two face features, the larger the value, the greater the difference between the two face feature vectors, the lower the similarity; is the value of the registered face feature vector in the i-th dimension, is the value of the face feature vector extracted from the person to be identified in the i-th dimension, is the dimension number of the feature vector;

[0025] If , then determine that the person to be identified is a registered personnel type, otherwise determine that the person to be identified is an unregistered personnel type.

[0026] Further, if identified as a registered personnel type, judge whether it is a temporary personnel by comparing the storage record, the expression is as follows:

[0027]

[0028] Indicates the judgment result, the judgment result includes and , indicates that the person is a temporary personnel, indicates that the person is a non-temporary personnel, indicates that the function is called to check whether the given identification is in the temporary personnel list, is an input parameter, representing the person corresponding to the minimum value of the Euclidean distance value

[0029] If it is a temporary person, get its stay time:

[0030]

[0031] where, is the stay time, is the current time, is the time when the person enters the community;

[0032] If exceeds the preset temporary person stay time, an alarm prompt is issued;

[0033] Unregistered personnel are identified by image acquisition equipment within the community and are assigned a temporary At the same time, the location information of the unregistered personnel is obtained and tracked, and the expression is as follows:

[0034]

[0035] where, is a variable for storing the obtained location information of the unregistered personnel, is a function call, which will query the current location of the unregistered personnel in the function and return the corresponding location information;

[0036] Send the location information to the community security personnel and control the image acquisition equipment for real-time tracking.

[0037] Further, real-time analysis of video streams is performed to monitor the bad behavior of registered personnel, and a behavior recognition algorithm is used for behavior prediction. If abnormal behavior is detected, the system automatically issues an alarm and records relevant image information for subsequent analysis;

[0038] Use the behavior recognition algorithm to analyze the video stream and detect potential bad behavior, and the expression is as follows:

[0039]

[0040] where, is the predicted behavior category, is the processed video stream, is a function call, which is used to identify the behavior patterns in and further identify and classify personnel behavior;

[0041] ​​If the bad behavior is detected, the system will issue an alarm and record the relevant image information.

[0042] Further, a self-learning mechanism is constructed, and the security personnel feedback on each alarm situation to update the database in time, and the expression is as follows:

[0043]

[0044] Wherein, is the feedback information of the security personnel on the alarm situation, This is a function that represents the update operation on the database.

[0045] The feedback information is specifically: alarm accuracy, the security personnel confirms whether the alarm is real, and provides relevant observation results; behavior description, the security personnel describes the specific event, the behavior and scene of the person involved; suggestions and improvements, the security personnel proposes improvement suggestions to help the system better identify and judge similar events;

[0046] Periodically retrain the model using newly collected data to improve recognition accuracy.

[0047] The application further provides an intelligent security information management device, which is used to execute the above-mentioned intelligent security information management method, comprising:

[0048] An image acquisition and detection module is used to acquire clear image data of key areas in the community, perform face detection on the acquired images, use a deep neural network to identify and extract face regions, apply a detection algorithm to each frame of image to obtain the coordinates of the face, and cut and store the face as an independent face image.

[0049] A feature extraction and comparison module is used to extract features from the extracted face image using a pre-trained deep learning model, and compare the extracted features with the stored registered database, calculate the Euclidean distance between the extracted feature vector and the registered vector, judge the type of personnel, and divide them into two categories: registered and unregistered.

[0050] An identification and alarm judgment module is used to classify image information according to the identification result, if the identification result is a registered personnel type, further judge whether it is a temporary personnel, and calculate the residence time, if the residence time exceeds the set threshold, an alarm is issued, and the location information of the unregistered personnel is obtained and tracked.

[0051] A behavior monitoring and prediction module is used to analyze video streams in real time, monitor the bad behavior of registered personnel, and use behavior recognition algorithms for behavior prediction, if abnormal behavior is detected, the system automatically issues an alarm and records the relevant image information for subsequent analysis.

[0052] The self-learning update mechanism module is used to build a self-learning mechanism, and the security personnel feed back on each alarm situation, update the database in a timely manner, and periodically retrain the model using new data to improve recognition accuracy and system adaptability.

[0053] The application also includes a computer-readable storage medium having a computer program stored thereon, wherein the computer program can implement the intelligent security information management method when executed by a processor.

[0054] Compared with the prior art, the application has the following beneficial effects:

[0055] The implementation of the intelligent security information management method not only improves the rapid recognition efficiency of personnel identity, significantly reduces the false alarm rate, but also enhances the monitoring ability of unregistered personnel, ensuring community safety. Through real-time behavior monitoring and abnormal behavior prediction, the system can timely issue an alarm to reduce potential security risks. At the same time, the introduction of the self-learning mechanism enables the security system to continuously improve and optimize according to feedback, improving the overall recognition accuracy and system adaptability, thereby effectively improving the overall level of intelligent security. In general, the scheme provides a more intelligent and dynamic solution for community security management, enhancing the work efficiency and response ability of security personnel. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The figure is a schematic diagram of the overall method of the application;

[0057] Figure 2 The figure is a schematic diagram of the device module of the application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below with reference to specific embodiments.

[0059] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the application should be understood as their usual meanings understood by those skilled in the art to which the application belongs. The terms "first", "second" and similar terms used in the application do not represent any order, quantity or importance, but are used to distinguish different components. The terms "include" or "contain" and similar terms mean that the elements or objects before the terms cover the elements or objects listed after the terms and their equivalents, without excluding other elements or objects. The terms "connect" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "up", "down", "left", "right" and the like only represent relative positional relationships, which may change accordingly when the absolute position of the described object changes.

[0060] Embodiment:

[0061] Please refer to Figure 1 The present application provides a technical solution:

[0062] The specific steps of the intelligent security information management method include:

[0063] Step 1: Collect clear image data of key areas in the community, detect human faces in the collected images, use deep neural networks to identify and extract face regions, apply detection algorithms to each frame of image to obtain face coordinates, and crop and store them as independent face images;

[0064] In this embodiment, the key areas include the community's gate, garage entrance, and community parks, squares, and children's playgrounds;

[0065] Get the image from the camera or video stream and pre-process it, including resizing and normalization, extract the face image from the collected image information, use the Caffe model provided by OpenCV for face detection;

[0066] For each frame of image Face region positioning, expression as follows:

[0067]

[0068] Where, is the current frame image, is the detected face region coordinates, specifically the top-left and bottom-right coordinates, is a function call, indicating the operation of face detection on the current frame image;

[0069] According to the detected coordinates, draw a rectangular frame on the original image to identify the face, and cut out the face region from the original image according to these coordinates and store it as an independent face image, denoted as .

[0070] The main advantage of step 1 is that face detection and extraction are performed through deep neural networks, which can significantly improve the accuracy and efficiency of face recognition. Compared with traditional rule-based methods, this step uses deep learning models that can automatically adapt to different environments and lighting conditions, maintaining stable recognition performance in complex scenarios. By processing each frame of image in real time, it can quickly obtain the face region and crop and store it, ensuring higher quality of basic data for subsequent recognition and reducing the probability of missed and false detections.

[0071] In the present patent scheme, step 1 can provide high-quality basic data for the overall system, ensuring the accuracy and effectiveness of subsequent steps. Accurate face detection and extraction lay a good foundation for the process of feature extraction and comparison, thereby improving the efficiency of the overall security system in personnel identification and classification. In addition, real-time processing capability enables the system to quickly respond to potential security threats, enhancing the community's security protection capability, creating favorable conditions for subsequent behavior monitoring and abnormal behavior detection, and improving the intelligent level of the entire security information management.

[0072] Step 2: Use the pre-trained deep learning model to extract the features of the extracted face image and compare it with the stored registered database. By calculating the Euclidean distance between the extracted feature vector and the registered vector, the type of personnel is determined, divided into registered and unregistered two categories;

[0073] In this embodiment, a pre-trained deep learning model is used for face feature extraction, and the expression is as follows:

[0074]

[0075] wherein, is the extracted face feature vector, containing various attributes of the face, is the cropped face image, represents calling the pre-trained deep learning model to perform forward propagation on to extract face features. The various attributes of the face are: facial contour, eye position, nose shape, mouth feature, and skin texture;

[0076] According to the following formula, the Euclidean distance between the extracted feature vector and the registered feature vector is calculated:

[0077]

[0078] wherein, is the Euclidean distance value, indicating the similarity of two face features. The larger the value, the greater the difference between the two face feature vectors, and the lower the similarity; is the value of the registered face feature vector in the i-th dimension, is the value of the face feature vector extracted from the to-be-identified personnel in the i-th dimension, is the dimension number of the feature vector;

[0079] If , then the to-be-identified personnel is determined to be a registered personnel type, otherwise the to-be-identified personnel is determined to be an unregistered personnel type.

[0080] In image recognition and face recognition, Euclidean distance is used to compare feature vectors to determine whether images belong to the same category. In multi-dimensional space, Euclidean distance can effectively measure the similarity between different points.

[0081] The main advantage of step 2 is the use of pre-trained deep learning models for face feature extraction and database comparison, which significantly improves the accuracy and speed of face recognition. Compared with traditional manual feature extraction methods, deep learning models can automatically learn and extract more complex and high-dimensional features, thereby improving the robustness to various face changes such as angle, lighting, expression, etc. In addition, by calculating the Euclidean distance between feature vectors for comparison, individual identity can be quickly and effectively determined, reducing the occurrence of false positives and false negatives.

[0082] In this patent scheme, step 2 can provide fast and accurate personnel classification results for the overall security system, ensuring the effectiveness of subsequent processing. By accurately identifying registered and unregistered personnel, the system can quickly respond to potential security threats and take appropriate measures such as issuing alarms and tracking unregistered personnel. This efficient identification mechanism not only enhances the security protection capabilities of the community, but also provides a reliable foundation for subsequent behavior monitoring and abnormal behavior detection, thereby promoting the intelligent and automated development of the entire security information management system.

[0083] Step 3: classify image information according to the recognition result, if it is identified as a registered personnel type, further judge whether it is a temporary personnel, and calculate its stay time, if it exceeds the set threshold, issue an alarm, at the same time, get the location information of unregistered personnel and track the processing;

[0084] In this embodiment, if it is identified as a registered personnel type, it is judged whether it is a temporary personnel by comparing the warehouse record, the expression is as follows:

[0085]

[0086] The judgment result includes and , indicates that the person is a temporary personnel, indicates that the person is a non-temporary personnel, indicates that the function is called to check whether the given identification is in the temporary personnel list, is an input parameter, indicating the person corresponding to the minimum Euclidean distance value;

[0087] If it is a temporary personnel, get its stay time:

[0088]

[0089] in, For the duration of stay, For the current time, The time for people to enter the community;

[0090] like If the preset stay time for temporary personnel is exceeded, an alarm will be issued.

[0091] Unregistered individuals were identified using image capture equipment within the community and assigned a temporary [position / unit]. Simultaneously, the location information of unregistered individuals is obtained and tracked, as shown in the following expression:

[0092]

[0093] in, It is a variable used to store the location information of unregistered persons. It's a function call; inside the function, it will... Query the current location of the unregistered person and return the corresponding location information;

[0094] Location information The message is sent to community security personnel, who then control the image acquisition equipment for real-time tracking.

[0095] The main advantage of step 3 lies in its ability to effectively improve the accuracy and response speed of community security management by monitoring and assessing the dwell time of registered individuals. Compared to traditional security monitoring methods that rely solely on video recording or manual patrols, which cannot monitor personnel behavior and dwell time in real time, potentially leading to overlooking security risks, this step automatically calculates dwell time and sets thresholds to achieve rapid identification of potential abnormal behavior, significantly enhancing the intelligence level of the security system.

[0096] In this patented solution, step 3 effectively enhances the security system's dynamic monitoring capabilities of personnel activities, ensuring effective management of all personnel within the community, especially temporary workers. By issuing timely alarms, the system can quickly respond to potential security threats, ensuring community safety. Simultaneously, tracking and processing the location information of unregistered personnel improves the accuracy of security warnings, promotes the intelligent and automated development of overall security information management methods, provides a solid foundation for subsequent behavior monitoring and anomaly detection, and ultimately enhances the overall security protection capabilities of the community.

[0097] Step 4: Analyze the video stream in real time, monitor the inappropriate behavior of registered personnel, use behavior recognition algorithms to predict behavior, and if abnormal behavior is detected, the system will automatically issue an alarm and record relevant image information for subsequent analysis.

[0098] In this embodiment, real-time video stream analysis is performed to monitor the undesirable behavior of registered personnel. Behavior recognition algorithms are used for behavior prediction. If abnormal behavior is detected, the system automatically sends an alarm and records relevant image information for subsequent analysis.

[0099] The video stream is analyzed using behavior recognition algorithms to detect potential undesirable behavior, expressed as follows:

[0100]

[0101] where, is the predicted behavior category, is the processed video stream, is a function call to identify behavior patterns, and further identify and classify personnel behavior.

[0102] If undesirable behavior is detected, the system will send an alarm and record relevant image information. The undesirable behavior includes fighting or violent behavior, damaging public property, theft, harassment, alcohol or drug abuse, and abnormal wandering behavior.

[0103] The logic for detecting undesirable behavior is as follows: a behavior recognition model based on convolutional neural network is constructed, multiple video data are obtained, and the video data are labeled as normal and undesirable behavior. The model is trained to obtain the trained behavior recognition model. The real-time acquired video data is input into the model to obtain the behavior label of the personnel.

[0104] The main advantage of step 4 is to analyze the video stream in real time to monitor the undesirable behavior of registered personnel. This method can efficiently and quickly identify potential security threats. Compared with traditional security methods relying on manual monitoring or post-viewing of videos, this step uses advanced behavior recognition algorithms to detect and predict abnormal behavior in real time, which helps to take timely measures and reduce the occurrence of security incidents. This automated monitoring method significantly improves the response speed and accuracy of the security system, ensuring the safety of the community.

[0105] In this patent solution, step 4 can provide more comprehensive and proactive security protection for the overall security system. By monitoring the behavior of registered personnel in real time, the system can not only timely detect and alarm abnormal situations, but also record relevant image information for subsequent analysis. This intelligent behavior detection mechanism enhances the community's early warning ability against undesirable behavior, thereby improving the safety of residents, providing data support for community security management, and promoting the continuous optimization and intelligent upgrading of the security system.

[0106] Step 5: Construct a self-learning mechanism, security personnel feedback on each alarm situation, update the database in a timely manner, at the same time, regularly use new data to retrain the model, improve the recognition accuracy and system adaptability;

[0107] In this embodiment, a self-learning mechanism is constructed, and security personnel feedback on each alarm situation to update the database in a timely manner. The expression is as follows:

[0108]

[0109] wherein, is the feedback information of the security personnel on the alarm situation, This is a function that represents the update operation on the database;

[0110] The feedback information is as follows: alarm accuracy, security personnel confirm whether the alarm is real and provide relevant observation results; behavior description, security personnel's specific description of the event and the behavior of the person involved; suggestions and improvements, security personnel's suggestions for improvement to help the system better identify and judge similar events;

[0111] Regularly use new collected data to retrain the model to improve recognition accuracy.

[0112] The main advantage of Step 5 is to construct a self-learning mechanism, so that the security system can continuously optimize and improve recognition accuracy through security personnel feedback. Compared with the prior art, traditional security systems often rely on static rules and fixed models, which are difficult to adapt to changing environments and needs. By collecting and analyzing feedback information from security personnel, the system can dynamically update the database and model, enhancing its adaptive ability, so that it can more effectively deal with various complex security threats in practical applications.

[0113] In this patent solution, Step 5 can significantly improve the intelligence level and practicality of the overall security system. Through continuous self-learning process, the system can effectively integrate historical data and user feedback, adjust recognition algorithms and strategies in a timely manner, and thus improve the prediction and handling ability of abnormal events. This feedback mechanism not only enhances the response speed and accuracy of the system, but also promotes the positive interaction between security personnel and technical systems, ultimately realizing the overall upgrade of community security management and providing residents with a safer and more reliable living environment.

[0114] Please refer to Figure 2 An intelligent security information management device, comprising:

[0115] An image acquisition and detection module is used to acquire clear image data of key areas in the community, to perform face detection on the acquired images, to use a deep neural network to identify and extract a face region, to obtain the coordinates of the face by applying a detection algorithm to each frame of image, and to crop and store the coordinates as an independent face image;

[0116] A feature extraction and comparison module is used to extract features from the face image using a pre-trained deep learning model, and to compare the extracted features with the stored registered database, to judge the type of personnel by calculating the Euclidean distance between the extracted feature vector and the registered vector, and to divide the personnel into two categories: registered and unregistered;

[0117] An identification and alarm judgment module is used to classify the image information according to the identification result, to further judge whether it is a temporary personnel if it is identified as a registered personnel type, and to calculate the stay time, and to issue an alarm if the stay time exceeds the set threshold, and to obtain the location information of the unregistered personnel and to track the unregistered personnel;

[0118] A behavior monitoring and prediction module is used to analyze the video stream in real time, to monitor the bad behavior of the registered personnel, and to use a behavior recognition algorithm to predict the behavior, and to automatically issue an alarm and record the relevant image information for subsequent analysis if an abnormal behavior is detected;

[0119] A self-learning update mechanism module is used to build a self-learning mechanism, to provide feedback on each alarm situation by security personnel, to update the database in a timely manner, and to periodically retrain the model using new data to improve the recognition accuracy and system adaptability.

[0120] The application also includes a computer-readable storage medium having a computer program stored thereon, wherein the computer program can implement the intelligent security information management method when executed by a processor.

[0121] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the nearest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0122] The above embodiments can be realized by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solutions.

[0123] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, and may be located in one place, or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment of the present application according to actual needs.

[0124] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for intelligent security information management, characterized in that, The specific steps include: Step 1: Collect clear image data of key areas in the community, perform face detection on the collected images, use a deep neural network to identify and extract the face region, apply a detection algorithm to each frame of image to obtain the coordinates of the face, and crop and store it as an independent face image; Step 2: Use a pre-trained deep learning model to extract features from the extracted face image and compare it with the stored registered database, calculate the Euclidean distance between the extracted feature vector and the registered vector to determine the type of personnel, and divide it into two categories: registered and unregistered; Step 3: According to the identification result, classify the image information, if it is identified as a registered personnel type, further judge whether it is a temporary personnel, and calculate the stay time, if it exceeds the set threshold, an alarm will be sent out, at the same time, the location information of the unregistered personnel is obtained and tracked; Step 4: Real-time analysis of video stream, monitoring the bad behavior of registered personnel, using behavior recognition algorithm for behavior prediction, if abnormal behavior is detected, the system automatically sends out an alarm and records the relevant image information for subsequent analysis; Step 5: Construct a self-learning mechanism, security personnel feedback on each alarm situation, update the database in time, at the same time, use new data to retrain the model regularly, improve the recognition accuracy and system adaptability; If it is identified as a registered personnel type, it is judged whether it is a temporary personnel by comparing the storage record, the expression is as follows: ； represents a judgment result, the judgment result including and , represents that the person is a temporary person, represents that the person is a non-temporary person, represents that a function is called to check whether the given identification is in a temporary person list, is an input parameter, representing a ; If it is a temporary personnel, get its stay time: ； wherein, is the time of stay, is the current time, is the time of entry of the person into the community; If If the preset temporary personnel stay time is exceeded, an alarm prompt is issued; Unregistered persons are identified by image capture equipment within the community and are assigned a temporary At the same time, the unregistered person's location information is obtained and tracking processing is performed, expressed as follows: ； wherein, is a variable for storing the acquired position information of the unregistered person, is a function call, and in the function, the current position of the unregistered person is queried and the corresponding position information is returned; the current position of the unregistered person is queried and the corresponding position information is returned; Location information The message was sent to community security personnel, who then controlled the image acquisition equipment for real-time tracking. Real-time analysis of video stream, monitoring the bad behavior of registered personnel, using behavior recognition algorithm for behavior prediction, if abnormal behavior is detected, the system automatically sends out an alarm and records the relevant image information for subsequent analysis; Use behavior recognition algorithm to analyze video stream, detect potential bad behavior, the expression is as follows: ； wherein, is a predicted class of behavior, is a processed video stream, is a function call to identify behavioral patterns in the video stream, and to identify and classify the behavior of the person. If abnormal behavior is detected, the system will send out an alarm and record the relevant image information.

2. The intelligent security information management method of claim 1, wherein: The key areas include the gates of the community, the entrances of the garages, and the parks, squares, and children's playgrounds in the community; Get the image from the camera or video stream, and preprocess the image, including resizing and normalization, extract the face image from the collected image information, use the Caffe model provided by OpenCV for face detection; For each frame of image Face region positioning is performed, and the expression is as follows: ； wherein, is a current frame image, is a detected face region coordinate, specifically, a left upper corner coordinate and a right lower corner coordinate, is a function call, indicating an operation of face detection on the current frame image; According to the detected coordinates, a rectangular frame is drawn on the original image to identify the face, and the face region is cropped from the original image according to the coordinates and stored as an independent face image, denoted as .

3. The intelligent security information management method according to claim 1, characterized in that: Use a pre-trained deep learning model to extract face features, the expression is as follows: ； wherein, is an extracted face feature vector, containing various attributes of the face, is a cropped face image, represents calling a pre-trained deep learning model on to perform forward propagation and extract face features; the various attributes of the face are specifically: facial contour, eye position, nose shape, mouth feature, and skin texture feature; According to the following formula, calculate the Euclidean distance between the extracted feature vector and the registered feature vector: ； wherein, is a Euclidean distance value, representing the similarity of two face feature vectors, the larger the value, the greater the difference between the two face feature vectors, the lower the similarity; is the value of the i-th dimension of the registered face feature vector, is the value of the i-th dimension of the face feature vector extracted from the person to be identified, is the dimension number of the feature vector; If then the person to be identified is determined to be of the registered person type, otherwise the person to be identified is determined to be of the unregistered person type.

4. The intelligent security information management method according to claim 1, characterized in that: Construct a self-learning mechanism, security personnel feedback on each alarm situation, update the database in time, the expression is as follows: ； wherein, is feedback information of the security personnel to the alarm situation, is a function representing an update operation on the database; The feedback information includes: alarm accuracy, security personnel confirm whether the alarm is real and provide relevant observation results; behavior description, security personnel describe the specific event and the behavior and scene of the involved personnel; suggestions and improvements, security personnel make improvement suggestions to help the system better identify and judge similar events; Periodically use new data for model retraining to improve recognition accuracy.

5. The intelligent security information management device, characterized in that: The intelligent security information management device is used for executing the intelligent security information management method in any one of claims 1-4, and comprises: An image acquisition and detection module is configured to acquire clear image data of key areas in the community, perform face detection on the acquired images, use a deep neural network to identify and extract face regions, apply a detection algorithm to each frame of image to obtain coordinates of the face, and cut and store the face as an independent face image. A feature extraction and comparison module is configured to use a pre-trained deep learning model to extract features of the extracted face image, compare the extracted features with stored registered database, calculate Euclidean distance between the extracted feature vector and the registered vector, judge the type of personnel, and divide the personnel into two categories of registered and unregistered. An identification and alarm judgment module is configured to classify image information according to the identification result, further judge whether the identified personnel type is temporary personnel if the personnel type is registered, calculate the stay time, and send an alarm if the stay time exceeds a set threshold, and obtain location information of the unregistered personnel and perform tracking processing. A behavior monitoring and prediction module is configured to analyze video streams in real time, monitor bad behaviors of the registered personnel, use a behavior recognition algorithm to perform behavior prediction, automatically send an alarm and record relevant image information for subsequent analysis if an abnormal behavior is detected. A self-learning update mechanism module is configured to build a self-learning mechanism, security personnel feed back on each alarm situation, update the database in a timely manner, periodically use new data to retrain the model, and improve recognition accuracy and system adaptability.

6. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program can implement the intelligent security information management method in any one of claims 1-4 when executed by a processor.

Images