Intelligent security system based on AI face recognition

Abstract

The invention relates to the technical field of intelligent security and protection, and discloses an intelligent security and protection system based on AI face recognition, which comprises the following modules: an image acquisition module used for acquiring image information recorded at a camera of current security and protection equipment, a face detection and recognition module used for carrying out portrait feature recognition, and a face recognition module used for carrying out face recognition. The system comprises a portrait recognition module, an image acquisition module, an accompanying movement module, a data storage and management module, a human face recognition module, an accompanying movement module, a human face recognition module and an accompanying movement management module, wherein the accompanying movement module controls a video recording device in the image acquisition module to move horizontally and vertically; the face recognition module is used for establishing, managing and storing a database of face data. According to the invention, through the built accompanying moving module, when it is detected that there is an abnormal person in the current area, the AI controls the accompanying driving module to drive the image recording device to carry out accompanying movement of the angle and the recording position along with the movement track of the abnormal person, so that the complete abnormal behavior of the abnormal person can be traced more comprehensively.

Description

Technical Field

[0001] The present invention relates to the field of intelligent security technology, and in particular to an intelligent security system based on AI face recognition. Background Art

[0002] Smart security refers to the use of advanced information technology, sensors, artificial intelligence and big data analysis to build a comprehensive security system. It achieves early warning and response to potential threats through real-time monitoring of the environment, data collection and analysis. In the smart security system, surveillance cameras, access control systems, alarm devices and other equipment are interconnected to form a unified security network.

[0003] This system can not only automatically identify suspicious activities, but also judge security risks through intelligent analysis, so as to take prompt measures. For example, when a surveillance camera detects an abnormal intrusion, it can immediately send an alarm to the relevant personnel, and automatically record and save the video for subsequent investigation. In addition, smart security systems are often combined with cloud computing to ensure remote storage and real-time access to data, making security management more efficient.

[0004] In the prior art, image monitoring equipment based on portrait recognition is often located at high places and can be manually controlled to adjust the security angle in real time and position it at any angle, thereby ensuring that the image monitoring equipment has fewer blind spots in monitoring and recognition during use.

[0005] However, in actual use, devices that manually adjust the monitoring angle usually lack real-time performance and require human intervention, which limits the ability to respond quickly. Therefore, when abnormal persons are identified through face recognition and their abnormal behavior needs to be traced, their complete abnormal behavior can only be traced through the combination of multiple groups of intelligent security devices in their monitoring areas. In view of this, we propose an intelligent security system based on AI face recognition. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides an intelligent security system based on AI face recognition, which solves the problem that when abnormal persons are identified through face recognition and their abnormal behavior needs to be traced, their complete abnormal behavior can only be traced through the combination and piecing together of multiple groups of intelligent security devices in their monitoring areas.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent security system based on AI face recognition, comprising the following modules:

[0008] Image acquisition module: used to obtain the image information captured by the current security device camera and the environmental information of the current camera recording area;

[0009] Face detection and recognition module: used to recognize portrait features based on the image data obtained in the image acquisition module, and to identify people based on the recognized portrait features;

[0010] Accompanying motion module: used to control the video recording device in the image acquisition module to perform horizontal and vertical plane movement and angle deflection;

[0011] Data storage and management module: used to establish and manage the database for storing face data, and establish a data indexing mechanism through a built-in search engine;

[0012] Alarm and response module: used to monitor and analyze the video stream provided by the image acquisition module in real time, and identify and analyze abnormal information and perform alarm processing after obtaining it;

[0013] System management and monitoring module: responsible for saving the internal parameters of each module and monitoring the performance indicators of the security system during equipment operation.

[0014] Preferably, the image acquisition module includes a camera hardware module, an environment perception module and an image preprocessing module. The camera hardware module is used to drive the operation of the video recording device and adjust and control the stability of the image during the process of acquiring image information. The environment perception module is used to detect light intensity, light direction and current camera operating temperature. The image processing module is used to perform grayscale processing, filtering and histogram equalization on the image information acquired in the camera hardware module.

[0015] Preferably, the face detection and recognition module includes a face detection module, a deep feature extraction module, a face recognition module and a liveness detection module. The face detection module is used to extract face feature information in the image, including facial texture, the position and shape of eyes, nose, mouth, and facial contour. The deep feature extraction module is used to extract local feature information when the current face is occluded or the person is in motion, and dynamically adjust the confidence threshold of face recognition based on real-time feedback from the scene, including the current light intensity, the person's movement speed and the degree of occlusion.

[0016] Preferably, the face recognition module is used to match the extracted facial features with known facial features in the database, and establish a real-time online learning mechanism to quickly incorporate new facial data through small sample incremental learning technology. The liveness detection module is used to detect whether the current portrait information is live by analyzing the portrait's biometric features including micro-expressions, blinking frequency, and head movements.

[0017] Preferably, the accompanying motion module includes an accompanying transformation module, a driving control module, a final positioning module and a secondary calibration module. The accompanying transformation module is used to plan the accompanying motion trajectory of the driving control module according to the motion trajectory of the face in the image acquisition module, and transmit the generated trajectory information to the driving control module. The driving control module is used to drive the image acquisition device to move according to the path information planned by the accompanying transformation module.

[0018] Preferably, the final positioning module is used to perform a secondary calibration on the final stop position of the image acquisition device, and to calibrate the motion parameters of the drive control module according to the deviation between the final stop position and the preset position; the secondary calibration module is used to fine-tune the current focal length, aperture size and focusing accuracy of the image acquisition device after it stops moving, and to record the adjusted parameters.

[0019] Preferably, the data storage and management module includes an AI storage optimization module and a face knowledge graph module. The AI ​​storage optimization module combines a deep learning algorithm with an image compression algorithm, performs targeted compression according to the characteristics of the face image, and reduces data storage space while ensuring the accuracy of face recognition. At the same time, the data of people who have not been accessed for a long time or have expired are automatically cleaned or archived to low-cost storage media. At the same time, the data is automatically organized and classified according to the importance and relevance of the data.

[0020] Preferably, the facial knowledge graph module is used to integrate facial data with relevant personnel information, including name, position, department, and social relationships to construct a knowledge graph. When querying and retrieving facial data, the relational reasoning ability of the knowledge graph is used to synchronously locate the historical records and related information of relevant personnel.

[0021] Preferably, the alarm and response module uses a reinforcement learning algorithm to train an alarm decision model. The model learns the best alarm method and timing according to different face recognition results, scene environment information and historical alarm data, including silent alarm or expelling abnormal personnel through sound and light alarm information.

[0022] Preferably, the system management and monitoring module includes a parameter configuration module and an AI diagnosis and optimization module. The parameter configuration module is used to configure the parameters of each module of the display system and provide a control page for adjusting various parameters in the system. The AI ​​diagnosis and optimization module uses a deep neural network to monitor and analyze the hardware and software operation data of the system in real time, and predicts possible faults in advance by learning the patterns of historical fault data.

[0023] The present invention provides an intelligent security system based on AI face recognition. It has the following beneficial effects:

[0024] 1. The accompanying mobile module established in the present invention can, when detecting the presence of abnormal persons in the current area, control the accompanying driving module through AI to drive the image recording device to move along the angle and recording position of the movement trajectory of the abnormal persons, so as to more comprehensively trace their complete abnormal behavior.

[0025] 2. The present invention establishes a deep feature extraction module, which can dynamically adjust the confidence threshold of face recognition based on real-time feedback from the scene, including the current light intensity, movement speed of people and degree of occlusion, thereby meeting the needs for portrait recognition in different environments.

[0026] 3. By combining the deep learning algorithm with the image compression algorithm, the present invention can perform AI intelligent compression according to the characteristics of the face image, and reduce data storage space while ensuring the accuracy of face recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is the module diagram of the intelligent security system based on AI face recognition;

[0028] Figure 2 It is a schematic diagram of the image acquisition module of the present invention;

[0029] Figure 3 A schematic diagram of a face detection and recognition module of the present invention;

[0030] Figure 4 It is a schematic diagram of the accompanying motion module of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0032] Example:

[0033] Please refer to the attached Figure 1 - Attachment Figure 4 , an embodiment of the present invention provides an intelligent security system based on AI face recognition, comprising the following modules:

[0034] Image acquisition module: used to obtain the image information captured by the current security device camera and the environmental information of the current camera recording area;

[0035] Face detection and recognition module: used to recognize portrait features based on the image data obtained in the image acquisition module, and to identify people based on the recognized portrait features;

[0036] Accompanying motion module: used to control the video recording device in the image acquisition module to perform horizontal and vertical plane movement and angle deflection;

[0037] Data storage and management module: used to establish and manage the database for storing face data, and establish a data indexing mechanism through a built-in search engine;

[0038] Alarm and response module: used to monitor and analyze the video stream provided by the image acquisition module in real time, and identify and analyze abnormal information and perform alarm processing after obtaining it;

[0039] System management and monitoring module: responsible for saving the internal parameters of each module and monitoring the performance indicators of the security system during equipment operation.

[0040] The image acquisition module includes a camera hardware module, an environment perception module and an image preprocessing module. The camera hardware module is used to drive the operation of the video recording device and adjust and control the stability of the image during the process of acquiring image information. The environment perception module is used to detect the light intensity, light direction and current camera operating temperature. The image processing module is used to perform grayscale processing, filtering and histogram equalization on the image information acquired in the camera hardware module.

[0041] The face detection and recognition module includes a face detection module, a deep feature extraction module, a face recognition module and a liveness detection module. The face detection module is used to extract face feature information in an image, including the position and shape of facial texture, eyes, nose, mouth, and facial contour. The deep feature extraction module is used to dynamically adjust the confidence threshold of face recognition based on the local feature information when the current face is blocked or the person is in motion, and based on the real-time feedback of the scene, including the current light intensity, the speed of the person's movement and the degree of occlusion;

[0042] For light intensity, assuming that it is measured by the average grayscale value of pixels, ranging from 0 to 255, the membership function of the "bright" fuzzy set can be defined as:

[0043]

[0044] The membership function of the “medium” fuzzy set is:

[0045]

[0046] The membership function of the “dark” fuzzy set is:

[0047]

[0048] For the speed of movement of personnel, assuming that it is measured by pixel displacement per second, the membership function of the "stationary" fuzzy set is:

[0049]

[0050] The membership function of the "slow" fuzzy set is:

[0051]

[0052] The membership function of the “fast” fuzzy set is:

[0053]

[0054] For the confidence threshold adjustment, assume that the initial confidence threshold is 0.8. The membership function of the "significantly reduced" fuzzy set is:

[0055]

[0056] The membership function of the “slightly reduced” fuzzy set is:

[0057]

[0058] The membership function of the “invariant” fuzzy set is:

[0059]

[0060] The membership function of the “slightly improved” fuzzy set is:

[0061]

[0062] The membership function of the “greatly improved” fuzzy set is:

[0063]

[0064] Create fuzzy rules:

[0065] For example, Rule 1: If the light intensity is "bright", the speed of the person's movement is "stationary", and the degree of occlusion is "none", then the confidence threshold adjustment amount is "slightly increased";

[0066] Rule 2: If the light intensity is "dark", the speed of movement is "fast", and the degree of occlusion is "partial", then the confidence threshold adjustment amount is "substantially reduced";

[0067] Perform fuzzy reasoning:

[0068] Assume that the average grayscale value of the pixels detected in real time is 80, the speed of movement is 7 pixels per second, and the degree of occlusion is the ratio of the occlusion area to the face area is 0.3.

[0069] First, calculate the membership of each input variable to the fuzzy set. For the light intensity x=80, μ 亮 (80) = 0;

[0070]

[0071] For the personnel movement speed x=7,μ 静止 (7)=0,μ 慢 (7)=0,μ 快 (7)=1.

[0072] For the occlusion degree x = 0.3, μ 无 (0.3)=0, μ 完全 (0.3) = 0

[0073] Reasoning is performed based on fuzzy rules. For Rule 2 (if the light intensity is "dark" and the speed of movement is "fast" and the degree of occlusion is "partial", then the confidence threshold adjustment is "substantially reduced"), the membership degree of the premise part is min(μ 暗 (80),μ 快 (7),μ 部分 (0.3))=min(0.67,1,0.4)=0.4.

[0074] Perform similar calculations on all rules to obtain the membership of the confidence threshold adjustment amount corresponding to each rule;

[0075] Defuzzification:

[0076] Assume that after fuzzy reasoning, the membership of "substantially reduced" is 0.4, the membership of "slightly reduced" is 0.3, the membership of "unchanged" is 0.2, the membership of "slightly increased" is 0.1, and the membership of "substantially increased" is 0;

[0077] According to the formula:

[0078]

[0079] Let the adjustment amount corresponding to "substantially reduced" be T 1 =-0.3, the adjustment amount corresponding to "slight reduction" is T 2 =-0.1, the adjustment amount corresponding to "unchanged" is T 3 =0, the adjustment amount corresponding to "slight increase" is T 4 =0.1, the adjustment amount corresponding to "substantially improved" is T5 =0.3;

[0080] but:

[0081]

[0082] Therefore, the confidence threshold was finally adjusted from the initial 0.8 to 0.8-0.14=0.66;

[0083] The face recognition module is used to match the extracted face features with the known face features in the database, and establish a real-time online learning mechanism to quickly incorporate new face data through small sample incremental learning technology. The liveness detection module is used to detect whether the current portrait information is live by analyzing the portrait's biometric features including micro-expressions, blinking frequency, head movements, etc.

[0084] The accompanying motion module includes an accompanying conversion module, a driving control module, a final positioning module and a secondary calibration module. The accompanying conversion module is used to plan the accompanying motion trajectory of the driving control module according to the motion trajectory of the face in the image acquisition module, and transmit the generated trajectory information to the driving control module. The driving control module is used to drive the image acquisition device to move according to the path information planned by the accompanying conversion module. The accompanying trajectory calculation for face recognition includes the following algorithm:

[0085] Let the target state vector x k (In this scenario, it can be the position, speed, and other information of the face in the image), and its state equation is:

[0086] x k =F k x k-1 +B k u k +w k

[0087] where F k is the state transfer matrix, which is used to describe the change of state from time k-1 to time k, B k is the control input matrix, u k is the control input vector (here related to the motion trend of the target, etc.), w k is process noise, assuming that its mean is 0 and its covariance is Q k Gaussian distribution of

[0088] The observation equation is:

[0089] z k =H k x k +v k

[0090] where zk is the observation vector (e.g., the face position information obtained by the image acquisition device), H k is the observation matrix, v k is the observation noise, with mean 0 and covariance R k Gaussian distribution

[0091] Prediction Step: A Priori Estimation A priori covariance estimates

[0092] Update step: Gain Posterior Estimation Posterior covariance Through these steps, the future position of the face can be predicted based on its historical motion trajectory, thereby planning the accompanying motion trajectory of the image acquisition device;

[0093] Example:

[0094] Assuming that you are tracking a face moving in a two-dimensional plane, the image acquisition device needs to follow the movement trajectory of the face.

[0095] Assume the face position coordinates in the two-dimensional plane are (x, y) and the horizontal and vertical velocities are v x ,v y , then the state vector x k =[x k ,y k ,v x,k ,v y,k ] T Here k represents the time;

[0096] Determine the state transfer matrix F k , control input matrix B k and the observation matrix H k :

[0097] For a simple uniform motion model (assuming that the face moves approximately uniformly in a short period of time), the state transfer matrix F k for:

[0098]

[0099] Where Δt is the time interval between two adjacent observations or controls. It means that within a time step, the position will change according to the speed, while the speed remains unchanged (uniform speed assumption).

[0100] Assuming there is no external control input (for simplicity in this example), the control input matrix B k =0;

[0101] The observation matrix H k It is used to map the state vector to the observation value (i.e. the position of the face in the image that can be observed by the image acquisition device). Assuming that only the position coordinates (x, y) of the face can be directly observed, then:

[0102]

[0103] Set the process noise covariance matrix Q k and the observation noise covariance matrix R k :

[0104] Process noise covariance matrix Q k Describes the uncertainty in the state transfer process. Assuming that there is a certain random fluctuation in face movement, such as the fluctuation range of speed in each direction is roughly the same, set as σ 2 ,but:

[0105]

[0106] Observation noise covariance matrix R k It reflects the error of the observation device (image acquisition device) in obtaining the face position information. Assume that the variance of the observation error in the x and y directions is but:

[0107]

[0108] Initialize the state estimate and covariance estimate:

[0109] When tracking begins, it is necessary to estimate the initial state of the face. Assuming that at the initial moment k = 0, the position of the face observed in the image is (x 0 ,y 0 ), but we first make a rough estimate of its speed and set it as (v x,0 ,v y,0 ), then the initial state estimate is:

[0110]

[0111] Initial covariance estimate P 0 It can be set according to the uncertainty of the initial state estimate. For example, if the initial estimate of the position is relatively accurate, but the accuracy of the velocity estimate is relatively low, it can be set to:

[0112]

[0113] in is the variance of the initial position estimate, is the variance of the initial velocity estimate;

[0114] Prediction Steps:

[0115] Assume that the time interval Δt = 1 second, the current time k = 1, and make a prediction based on the parameters set previously. Because B 1 =0, so F 1 (Substituting Δt=1 in the form defined above) and Substituting the value of into the calculation, we can get the prior estimated state vector;

[0116] A priori covariance estimates Similarly, substitute the corresponding values ​​for calculation;

[0117] Update steps:

[0118] Assume that at time k = 1, the position of the face observed by the image acquisition device is (z 1x ,z 1y ), then the observation vector z 1 =[z 1x ,z 1y ] T ;

[0119] First calculate the gain Then the posterior estimate and the posterior covariance By repeating the above prediction and updating steps, as time goes by, the accompanying motion trajectory of the image acquisition device can be more accurately planned according to the motion trajectory of the face;

[0120] The final positioning module is used to perform secondary calibration on the final stop position of the image acquisition device, and to calibrate the motion parameters of the drive control module according to the deviation between the final stop position and the preset position. The secondary calibration module is used to fine-tune the current focal length, aperture size and focus accuracy of the image acquisition device after it stops moving, and to record the adjusted parameters.

[0121] The data storage and management module includes an AI storage optimization module and a face knowledge graph module. The AI ​​storage optimization module combines the deep learning algorithm with the image compression algorithm, performs targeted compression according to the characteristics of the face image, and reduces the data storage space under the premise of ensuring the accuracy of face recognition. At the same time, the data of personnel that have not been accessed for a long time or have expired are automatically cleaned or archived to low-cost storage media. At the same time, the data is automatically sorted and classified according to the importance and relevance of the data. The deep learning algorithm and image compression algorithm used in the process are as follows:

[0122] Autoencoder algorithm formula: An autoencoder consists of an encoder and a decoder. Let the input face image be x, the encoder function be f, and the decoder function be g. The encoder maps the input x to the hidden representation z = f(x);

[0123] For example, in a simple fully connected autoencoder, if the input layer has n neurons and the hidden layer has m neurons (m < n), for the input vector x = [x 1 , x 2 , …, x n , the output of the hidden layer z = [z 1 , z 2 , …, z m , where w ij is the weight, b i is the bias, and f i is the activation function (such as the sigmoid function );

[0124] The decoder reconstructs the hidden representation z into For example, under the same fully connected structure, the reconstructed output where v kl is the decoder weight, c k is the decoder bias, and g k is the activation function;

[0125] The goal of training the autoencoder is to minimize the reconstruction loss

[0126] Image compression algorithm:

[0127] The compression process mainly includes discrete cosine transform (DCT), quantization, and entropy coding. For an 8×8 image block I(x, y), the discrete cosine transform formula is:

[0128]

[0129] where C(u) and C(v) are coefficients. When u = v = 0, otherwise C(u) = C(v) = 1;

[0130] Quantization is to divide the DCT coefficients by the values in the quantization table and round. The quantized coefficients S(u, v) is the quantization step in the quantization table;

[0131] Entropy coding is to encode the quantized coefficients to further reduce the storage space. When combined with the autoencoder, the DCT coefficients can be quantized in a targeted manner according to the features extracted by the autoencoder. For example, for the DCT coefficient area corresponding to the features that the autoencoder considers important, a smaller quantization step size is used, and for the secondary feature area, a larger quantization step size is used;

[0132] The face knowledge graph module is used to integrate face data with related personnel information, including name, position, department, and social relationships to construct a knowledge graph. When querying and retrieving face data, the relational reasoning ability of the knowledge graph is used to synchronously locate the historical records and related information of related personnel.

[0133] The alarm and response module uses a reinforcement learning algorithm to train an alarm decision model. The model learns the best alarm method and timing based on different face recognition results, scene environment information and historical alarm data, including silent alarm or expelling abnormal personnel through sound and light alarm information.

[0134] The system management and monitoring module includes a parameter configuration module and an AI diagnosis and optimization module. The parameter configuration module is used to configure and display the parameters of each module of the system and provide a control page for adjusting various parameters in the system. The AI ​​diagnosis and optimization module uses a deep neural network to monitor and analyze the system's hardware and software operation data in real time, and predicts possible faults in advance by learning the patterns of historical fault data.

[0135] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent security system based on AI face recognition, characterized in that: Includes the following modules, Image acquisition module: used to obtain the image information captured by the current security device camera and the environmental information of the current camera recording area; Face detection and recognition module: used to recognize portrait features based on the image data obtained in the image acquisition module, and to identify people based on the recognized portrait features; Accompanying motion module: used to control the video recording device in the image acquisition module to perform horizontal and vertical plane movement and angle deflection; Data storage and management module: used to establish and manage the database for storing face data, and establish a data indexing mechanism through a built-in search engine; Alarm and response module: used to monitor and analyze the video stream provided by the image acquisition module in real time, and identify and analyze abnormal information and perform alarm processing after obtaining it; System management and monitoring module: responsible for saving the internal parameters of each module and monitoring the performance indicators of the security system during equipment operation.

2. According to claim 1, the intelligent security system based on AI face recognition is characterized in that: The image acquisition module includes a camera hardware module, an environment perception module and an image preprocessing module. The camera hardware module is used to drive the operation of the video recording device and adjust and control the stability of the image during the process of acquiring image information. The environment perception module is used to detect the light intensity, light direction and the current camera operating temperature. The image processing module is used to perform grayscale processing, filtering and histogram equalization on the image information acquired in the camera hardware module.

3. The intelligent security system based on AI face recognition according to claim 1, characterized in that: The face detection and recognition module includes a face detection module, a deep feature extraction module, a face recognition module and a liveness detection module. The face detection module is used to extract face feature information in the image, including facial texture, the position and shape of eyes, nose, mouth, and facial contour. The deep feature extraction module is used to extract local feature information when the current face is occluded or the person is in motion, and based on real-time feedback from the scene, including the current light intensity, the person's movement speed and the degree of occlusion, dynamically adjust the confidence threshold of face recognition.

4. The intelligent security system based on AI face recognition according to claim 3 is characterized in that: The face recognition module is used to match the extracted face features with the known face features in the database, and establish a real-time online learning mechanism to quickly incorporate new face data through small sample incremental learning technology. The liveness detection module is used to detect whether the current portrait information is live by analyzing the portrait's biometric features including micro-expressions, blinking frequency, and head movements.

5. The intelligent security system based on AI face recognition according to claim 1, characterized in that: The accompanying motion module includes an accompanying conversion module, a driving control module, a final positioning module and a secondary calibration module. The accompanying conversion module is used to plan the accompanying motion trajectory of the driving control module according to the motion trajectory of the face in the image acquisition module, and transmit the generated trajectory information to the driving control module. The driving control module is used to drive the image acquisition device to move according to the path information planned by the accompanying conversion module.

6. The intelligent security system based on AI face recognition according to claim 5, characterized in that: The final positioning module is used to perform a secondary calibration on the final stop position of the image acquisition device, and to calibrate the motion parameters of the drive control module according to the deviation between the final stop position and the preset position. The secondary calibration module is used to fine-tune the current focal length, aperture size and focus accuracy of the image acquisition device after it stops moving, and to record the adjusted parameters.

7. The intelligent security system based on AI face recognition according to claim 1, characterized in that: The data storage and management module includes an AI storage optimization module and a face knowledge graph module. The AI ​​storage optimization module combines a deep learning algorithm with an image compression algorithm to perform targeted compression based on the characteristics of the face image and reduce data storage space while ensuring the accuracy of face recognition. At the same time, the data of people who have not been accessed for a long time or have expired are automatically cleaned or archived to low-cost storage media. At the same time, the data is automatically organized and classified according to its importance and relevance.

8. The intelligent security system based on AI face recognition according to claim 7, characterized in that: The face knowledge graph module is used to integrate face data with related personnel information, including name, position, department, and social relationships to construct a knowledge graph. When querying and retrieving face data, the relational reasoning ability of the knowledge graph is used to synchronously locate the historical records and related information of related personnel.

9. The intelligent security system based on AI face recognition according to claim 1, characterized in that: The alarm and response module uses a reinforcement learning algorithm to train an alarm decision model. The model learns the best alarm method and timing based on different face recognition results, scene environment information and historical alarm data, including silent alarm or expelling abnormal personnel through sound and light alarm information.

10. The intelligent security system based on AI face recognition according to claim 1, characterized in that: The system management and monitoring module includes a parameter configuration module and an AI diagnosis and optimization module. The parameter configuration module is used to configure and display the parameters of each module of the system and provide a control page for adjusting various parameters in the system. The AI ​​diagnosis and optimization module uses a deep neural network to monitor and analyze the system's hardware and software operation data in real time, and predicts possible faults in advance by learning the patterns of historical fault data.

Images