An intelligent visual big data access control system

Abstract

The present invention relates to the field of face recognition technology, and discloses an intelligent visual big data access control system, comprising a data entry module, a face acquisition module, a face image preprocessing module, a face extraction module, a face detection module, a background detection module, a recognition result processing module, an early warning module, an instruction execution module and an access control terminal; by providing the face detection module, the background detection module and the recognition result processing module, it is beneficial to verify and identify the face to be identified and the background face, and further analyze according to the recognition result, so as to send a corresponding early warning instruction, detect the stranger face appearing in the background, if the tailing phenomenon occurs, timely early warning and emergency call can be made, and effective prompts are given; if the stranger face appearing in the background is only an unrecorded face, it will not affect the normal access control; if the stranger face appearing in the background is confirmed to be a stranger, the safety of girls living alone can be guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of face recognition technology, and more specifically to an intelligent visual big data access control system. Background Art

[0002] Intelligent vision technology is one of the core technologies in the security field. The so-called intelligent vision technology is to automatically analyze and process the collected video through computers, capture the targets of interest in it, and further obtain the target's appearance time, movement trajectory, color and other information. By analyzing the above information of each target, it can find the dangers, violations or suspicious targets in the video, and perform real-time alarms, early warnings, storage and post-retrieval of these behaviors and targets; intelligent vision technology can be further divided into several application directions according to the adopted technology and application fields, such as intelligent video surveillance, intelligent video retrieval, intelligent traffic monitoring and face recognition technology.

[0003] The existing access control system recognizes the face and compares the face with the pre-entered face features. If the recognition is correct, the face recognition is successful. If the recognition is wrong, the face recognition fails. This verifies whether the user's identity is legal and ensures the security of the access control system and the accuracy of personnel identification management.

[0004] However, the existing access control system cannot detect the background technology in facial recognition, cannot detect the risks behind, cannot detect and warn in advance of possible stalking, and cannot ensure the safety of girls living alone. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides an intelligent visual big data access control system to solve the problems existing in the above-mentioned background technology.

[0006] The present invention provides the following technical solutions: an intelligent visual big data access control system, comprising a data entry module, a face acquisition module, a face image preprocessing module, a face extraction module, a face detection module, a background detection module, a recognition result processing module, an early warning module, an instruction execution module and an access control terminal;

[0007] The data entry module is used to enter faces in advance and establish a face recognition database;

[0008] The face acquisition module is used to acquire the face image to be identified and transmit it to the background extraction module;

[0009] The face image preprocessing module is used to preprocess the face image collected by the face collection module and transmit the preprocessed face image to the face extraction module;

[0010] The face extraction module extracts faces from face images based on a convolutional neural network, can extract multiple faces at the same time, distinguish faces to be identified from background faces, transmit the face data to be identified to the face detection module, and transmit the background face data to the background detection module;

[0011] The face detection module is used to verify and identify the face to be identified, and transmit the identification result to the identification result processing module;

[0012] The background detection module is used to verify and identify the background face and transmit the identification result to the identification result processing module;

[0013] The recognition result processing module is used to receive the recognition results of the face detection module and the background detection module for analysis, and send corresponding warning instructions to the warning module according to the analysis results;

[0014] The warning module is used to receive the warning instruction sent by the recognition result processing module, send a warning prompt to the access control terminal and receive feedback information from the access control terminal;

[0015] The instruction execution module is used to execute the access control release instruction sent by the recognition result processing module and release the access control;

[0016] The access control terminal is a display terminal used for human-computer interaction and also feeds back information to the early warning module.

[0017] Preferably, the specific manner in which the data entry module establishes the face recognition database is as follows:

[0018] Collect n images of the same face as input face data, the image size is A×B, and the input face data is represented by a matrix as follows: ; where x i The result of matrix vectorization of the i-th face image; if the i-th image is represented as , after matrix vectorization, the resulting matrix is ​​expressed as ; Calculate the average value of the relative pixels of n images of the same face to obtain the smooth structure of the face as a reference for identifying the face image.

[0019] Preferably, the calculation formula of the average value is: ; Where δ is the average value; i=1, 2, 3...n;

[0020] Based on the calculated average value, the covariance matrix of the input face data is constructed, and the construction formula is: ; where d i represents the corresponding eigenvector value, ; N is the original face image, and ; M represents the covariance matrix; T represents the "vectorization" of the image eigenvalues, and all eigenvector values ​​form the eigenface space m; the averaged data is projected in the eigenface space ; Among them, p i is the projection in the eigenface space, i=1, 2, 3...n.

[0021] Preferably, the convolutional neural network of the face extraction module includes a convolution layer, a pooling layer, a dropout layer, a tiling layer, a fully connected layer and a classification layer;

[0022] The convolution layer receives the face image preprocessed by the face image preprocessing module to obtain input image data , where y1 is the input image, W1 is the width of the input image, H1 is the height of the input image, and D is the number of channels; if the image is RGB type, the number of channels is 3; if the image is a grayscale image, the number of channels is 1; the input image data is obtained after the convolution layer operation to obtain the output image data ; Where y2 is the output image, W2 is the width of the output image, and H2 is the height of the output image; ; ; Where F1 is the width and height of the convolution kernel, S1 is the step size of each convolution operation, and P is the padding number;

[0023] The pooling layer receives the data y2 after the convolution operation, and outputs the image data after the pooling operation: ,in, , , W3 is the width of the pooling layer output image, H3 is the height of the pooling layer output image, F2 is the width and height of the pooling kernel, and S2 is the step size of each pooling operation;

[0024] The calculation formula of the Dropout layer is: ; ;Wherein, the Bernoulli function represents the generation probability r vector, i.e., a vector of 0 and 1; Represents the data generated after the input data y3 is dropped;

[0025] The tiling layer transforms the input data into one dimension. Convert to , the flat layer is the transition layer from convolution, pooling and Dropout to the fully connected layer;

[0026] The fully connected layer finally obtains the output data , where y5 represents the output data of the fully connected layer, ω is the weight, y4 is the input data, and b represents the offset;

[0027] The classification layer adopts the softmax classification function, and the formula is expressed as: ,in, represents the output probability, z j represents the input data of each neuron in y5, and J represents the number of all neurons.

[0028] Preferably, the face detection module verifies and identifies the face to be identified, including the following steps:

[0029] Step S01: Noise filtering is performed on the image of the face to be recognized. The formula is: , where f(x,y) represents the processed face image pixels to be recognized, g(xv,yt) represents the corresponding input face image pixels to be recognized, and W represents 3×3 and 5×5 two-dimensional templates;

[0030] Step S02: The face to be identified in step S01 is averaged and marked as τ, and the face image to be identified is projected into the feature face space m. The feature vector of the face to be identified is expressed as , where ρ τ is the feature vector of the face to be identified, δ is the average value in the face recognition database; calculate the vector distance between the feature vector of the face to be identified and the feature face space of the face recognition database: ,in, is the vector distance between the feature vector of the face to be identified and the feature face space of the face recognition database; p i is the projection in the eigenface space, i=1, 2, 3...n; n is the number of images of the same face;

[0031] Step S03: Perform threshold comparison on the vector distance in step S02. If , indicating that the face image to be recognized matches the face image entered into the face recognition database, and the recognition is successful; otherwise, it indicates that the face image to be recognized does not match the face image entered into the face recognition database, and the recognition fails; where YU is the threshold.

[0032] Preferably, the specific method of the background detection module to verify and identify the background face is:

[0033] Extracting features of background faces , where N is the N dimension, extracting facial features from the face recognition database , where q is the qth face entered; calculate the Euclidean distance between the features of the background face and the features of the face in the face recognition database: , where Dis q is the Euclidean distance between the background face feature and the qth face feature entered into the face recognition database;

[0034] If there exists a q such that Dis q <UI, it indicates that the background face image matches the qth face image in the face recognition database and the recognition is successful; otherwise, it indicates that the background face image does not match any face in the face recognition database and the recognition fails; the UI is a preset background threshold.

[0035] Preferably, the specific manner in which the recognition result processing module analyzes the recognition results of the face detection module and the background detection module is:

[0036] If the recognition results of the face detection module and the background detection module are both successful, the analysis result is "successful", and the access control release command is sent to the command execution module;

[0037] If the recognition result of the face detection module is successful, and the recognition result of the background detection module is failed, the analysis result is "there is a hidden danger of unfamiliarity", and an early warning instruction is sent to the early warning instruction module;

[0038] If the recognition result of the face detection module is failure, no matter whether the recognition result of the background detection module is success or failure, the analysis result is "failure". At this time, the access control release command is not sent to the command execution module, and the access control cannot be released.

[0039] Technical effects and advantages of the present invention:

[0040] The present invention is provided with a face detection module, a background detection module and a recognition result processing module, which is conducive to verifying and identifying the face to be identified and the background face, and further analyzing according to the recognition result, so as to send a corresponding early warning instruction, detect the stranger face appearing in the background, and if a tailing phenomenon occurs, a warning and emergency call can be issued in time to provide effective prompts. If the stranger face appearing in the background is only an unrecorded face, it will not affect the normal access control entry and exit. If the stranger face appearing in the background is confirmed to be a stranger, the safety of girls living alone can be guaranteed, and an emergency call can be automatically made without receiving feedback from the access control terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a structural diagram of the intelligent visual big data access control system of the present invention. DETAILED DESCRIPTION

[0042] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are merely illustrative. The intelligent visual big data access control system involved in the present invention is not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0043] The present invention provides an intelligent visual big data access control system, including a data entry module, a face acquisition module, a face image preprocessing module, a face extraction module, a face detection module, a background detection module, a recognition result processing module, an early warning module, an instruction execution module and an access control terminal;

[0044] The data entry module is used to enter faces in advance and establish a face recognition database;

[0045] The face acquisition module is used to acquire the face image to be identified and transmit it to the background extraction module; the face image to be identified can be acquired through the camera of the access control system;

[0046] The face image preprocessing module is used to preprocess the face image collected by the face collection module and transmit the preprocessed face image to the face extraction module;

[0047] The face extraction module extracts faces from face images based on a convolutional neural network, can extract multiple faces at the same time, distinguish faces to be identified from background faces, transmit the face data to be identified to the face detection module, and transmit the background face data to the background detection module;

[0048] The face detection module is used to verify and identify the face to be identified, and transmit the identification result to the identification result processing module;

[0049] The background detection module is used to verify and identify the background face and transmit the identification result to the identification result processing module;

[0050] The recognition result processing module is used to receive the recognition results of the face detection module and the background detection module for analysis, and send corresponding warning instructions to the warning module according to the analysis results;

[0051] The warning module is used to receive the warning instruction sent by the recognition result processing module, send a warning prompt to the access control terminal and receive feedback information from the access control terminal;

[0052] The instruction execution module is used to execute the access control release instruction sent by the recognition result processing module and release the access control;

[0053] The access control terminal is a display terminal, which is used for human-computer interaction and can also feed back information to the early warning module. The display terminal can be a small program or an app. After receiving the early warning prompt, the access control terminal will feed back the content of the early warning prompt, and the content of the early warning prompt is "The background face is detected and not recorded". You can click "Ignore" or "Warning". "Ignore" means that the background face is not a stranger's face, but just a face that has not been recorded; "Warning" means that the background face is a stranger's face; and the click content will be fed back to the early warning module. If the feedback received by the early warning module is "Warning", an emergency call will be made; if the early warning module does not receive feedback from the access control terminal within a certain period of time, an emergency call will be automatically made; the certain period of time can be set in advance through the display terminal, and can be 5 minutes, 15 minutes, 30 minutes, etc. The emergency call object can also be set in advance through the display terminal.

[0054] In this embodiment, it should be specifically explained that the specific manner in which the data entry module establishes the face recognition database is as follows:

[0055] Collect n images of the same face as input face data, the image size is A×B, and the input face data is represented by a matrix as follows: ; where x i The result of matrix vectorization of the i-th face image; for example, if the i-th image is represented as , after matrix vectorization, the resulting matrix is ​​expressed as ; Calculate the average value of the relative pixels of n images of the same face to obtain the smooth structure of the face as a reference for identifying the face image;

[0056] The calculation formula of the average value is: ; Where δ is the average value; i=1, 2, 3...n;

[0057] Based on the calculated average value, the covariance matrix of the input face data is constructed, and the construction formula is: ; where d i represents the corresponding eigenvector value, ; N is the original face image, and ; M represents the covariance matrix; T represents the "vectorization" of the image eigenvalues, and all eigenvector values ​​form the eigenface space m; the averaged data is projected in the eigenface space ; Among them, p i is the projection in the eigenface space, i=1, 2, 3, ... n;

[0058] The data entry module can enter multiple face images. This embodiment only lists the method for establishing a recognition database for one face. If multiple faces are entered in the same way, a recognition database for all entered faces can be established.

[0059] In this embodiment, it should be specifically explained that the convolutional neural network of the face extraction module includes a convolution layer, a pooling layer, a dropout layer, a tiling layer, a fully connected layer and a classification layer;

[0060] The convolution layer receives the face image preprocessed by the face image preprocessing module to obtain input image data , where y1 is the input image, W1 is the width of the input image, H1 is the height of the input image, and D is the number of channels; if the image is RGB type, the number of channels is 3; if the image is a grayscale image, the number of channels is 1; the input image data is obtained after the convolution layer operation to obtain the output image data ; Where y2 is the output image, W2 is the width of the output image, and H2 is the height of the output image; ; ; Where F1 is the width and height of the convolution kernel, S1 is the step size of each convolution operation, and P is the padding number;

[0061] The pooling layer receives the data y2 after the convolution operation, and outputs the image data after the pooling operation: ,in, , , W3 is the width of the pooling layer output image, H3 is the height of the pooling layer output image, F2 is the width and height of the pooling kernel, and S2 is the step size of each pooling operation; the pooling operation is divided into mean pooling and maximum pooling, and the most commonly used one is maximum pooling, the purpose is to reduce spatial information, improve computational efficiency, and prevent overfitting;

[0062] The calculation formula of the Dropout layer is: ; ;Wherein, the Bernoulli function represents the generation probability r vector, i.e., a vector of 0 and 1; Represents the data generated by the input data y3 after dropout; the Dropout layer can effectively alleviate the occurrence of overfitting and achieve the effect of regularization to a certain extent;

[0063] The tiling layer transforms the input data into one dimension. Convert to , this layer is the transition layer from convolution, pooling and Dropout to the fully connected layer;

[0064] The fully connected layer uses the one-dimensional data obtained in the above layer as the input data of the fully connected layer, performs weighted operations and offset processing on the one-dimensional data, and finally obtains the output data , where y5 represents the output data of the fully connected layer, ω is the weight, y4 is the input data, and b represents the offset;

[0065] The classification layer uses the softmax classification function, also known as the normalized exponential function, which is suitable for multi-face recognition; the formula is expressed as:

[0066] ,in, represents the output probability, z j represents the input data of each neuron in y5, and J represents the number of all neurons.

[0067] In this embodiment, it should be specifically explained that after the face extraction module recognizes the face target through the convolutional neural network, the recognized face target is marked as L1, L2, L3...L u Each face target corresponds to a target box, and the area of ​​the target box is used as the face target area. The face target with the largest area is used as the face to be identified and transmitted to the face detection module; the rest are recorded as background faces and transmitted to the background detection module.

[0068] In this embodiment, it should be specifically explained that the face detection module verifies and identifies the face to be identified by the following steps:

[0069] Step S01: Noise filtering is performed on the image of the face to be recognized. The formula is: , where f(x,y) represents the processed face image pixels to be recognized, g(xv,yt) represents the corresponding input face image pixels to be recognized, and W represents 3×3 and 5×5 two-dimensional templates;

[0070] The noise filtering adopts the noise suppression method of value filtering, and sets the grayscale value of the pixel of the face image to be recognized as the termination filter of all the pixels in the face recognition database, thereby suppressing the image noise; firstly, the difference between the grayscale value of a certain pixel and the surrounding noise pixels is calculated, and the noise with a larger difference with the pixel is defined as the intermediate value caused by filtering, so as to avoid the problem of blurred image details during linear filtering;

[0071] Step S02: The face to be identified in step S01 is averaged and marked as τ, and the face image to be identified is projected into the feature face space m. The feature vector of the face to be identified is expressed as , where ρ τis the feature vector of the face to be identified, δ is the average value in the face recognition database; calculate the vector distance between the feature vector of the face to be identified and the feature face space of the face recognition database: ,in, is the vector distance between the feature vector of the face to be identified and the feature face space of the face recognition database;

[0072] Step S03: Perform threshold comparison on the vector distance in step S02. If , indicating that the face image to be recognized matches the face image entered in the face recognition database, and the recognition is successful; otherwise, it indicates that the face image to be recognized does not match the face image entered in the face recognition database, and the recognition fails; the YU is the threshold, and the calculation formula is: , where a=1, 2, 3…n, b=1, 2, 3…n.

[0073] In this embodiment, it should be specifically explained that the specific method in which the background detection module verifies and identifies the background face is:

[0074] Extracting features of background faces , where N is the N dimension, extracting facial features from the face recognition database , where q is the qth face entered; calculate the Euclidean distance between the features of the background face and the features of the face in the face recognition database: , where Dis q is the Euclidean distance between the background face feature and the qth face feature entered into the face recognition database;

[0075] If there exists a q such that Dis q <UI, it indicates that the background face image matches the qth face image in the face recognition database and the recognition is successful; otherwise, it indicates that the background face image does not match any face in the face recognition database and the recognition fails; the UI is a preset background threshold, and the value of the UI can be specifically set by those skilled in the art according to actual conditions, and this embodiment does not specifically limit this value.

[0076] In this embodiment, it should be specifically explained that the specific manner in which the recognition result processing module analyzes the recognition results of the face detection module and the background detection module is:

[0077] If the recognition results of the face detection module and the background detection module are both successful, the analysis result is "success", and the access control release command is sent to the command execution module; the faces that appear at this time are all face images that have been entered into the face database in advance, so there are no stranger faces, the environment is safe, and the access control can be released;

[0078] If the recognition result of the face detection module is successful, and the recognition result of the background detection module is failed, the analysis result is "there is a hidden danger of strangers", and an early warning instruction is sent to the early warning instruction module; at this time, an unrecorded face appears in the background, and there may be a situation of being followed by a stranger. At this time, a warning instruction needs to be sent to confirm whether the background face is a stranger's face;

[0079] If the recognition result of the face detection module is failure, no matter whether the recognition result of the background detection module is success or failure, the analysis result is "failure". At this time, the access control release command is not sent to the command execution module, and the access control cannot be released.

[0080] In the present embodiment, it should be specifically explained that the difference between the present embodiment and the prior art mainly lies in that the present embodiment is equipped with a face detection module, a background detection module and a recognition result processing module. By verifying and identifying the face to be identified and the background face, and further analyzing the recognition result, a corresponding early warning instruction is sent to detect the stranger face appearing in the background. If a stalking phenomenon occurs, an early warning and emergency call can be made in time to provide effective prompts. If the stranger face appearing in the background is only an unrecorded face, it will not affect the normal access control entry and exit. If the stranger face appearing in the background is confirmed to be a stranger, the safety of girls living alone can be guaranteed, and an emergency call can be made automatically without receiving feedback from the access control terminal.

[0081] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0082] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An intelligent visual big data access control system, characterized by: It includes data entry module, face acquisition module, face image preprocessing module, face extraction module, face detection module, background detection module, recognition result processing module, early warning module, instruction execution module and access control terminal; The data entry module is used to enter faces in advance and establish a face recognition database; The face acquisition module is used to acquire the face image to be identified and transmit it to the background extraction module; The face image preprocessing module is used to preprocess the face image collected by the face collection module and transmit the preprocessed face image to the face extraction module; The face extraction module extracts faces from face images based on a convolutional neural network, can extract multiple faces at the same time, distinguish faces to be identified from background faces, transmit the face data to be identified to the face detection module, and transmit the background face data to the background detection module; The face detection module is used to verify and identify the face to be identified, and transmit the identification result to the identification result processing module; The background detection module is used to verify and identify the background face and transmit the identification result to the identification result processing module; The recognition result processing module is used to receive the recognition results of the face detection module and the background detection module for analysis, and send corresponding warning instructions to the warning module according to the analysis results; The warning module is used to receive the warning instruction sent by the recognition result processing module, send a warning prompt to the access control terminal and receive feedback information from the access control terminal; The instruction execution module is used to execute the access control release instruction sent by the recognition result processing module and release the access control; The access control terminal is a display terminal used for human-computer interaction and also feeds back information to the early warning module.

2. According to claim 1, the intelligent visual big data access control system is characterized by: The specific method of establishing the face recognition database by the data entry module is as follows: Collect n images of the same face as input face data, the image size is A×B, and the input face data is represented by a matrix as follows: ; where x i The result of matrix vectorization of the i-th face image; if the i-th image is represented as , after matrix vectorization, the resulting matrix is ​​expressed as ; Calculate the average value of the relative pixels of n images of the same face to obtain the smooth structure of the face as a reference for identifying the face image.

3. The intelligent visual big data access control system according to claim 2 is characterized by: The calculation formula of the average value is: ; Where δ is the average value; i=1, 2, 3...n; Based on the calculated average value, the covariance matrix of the input face data is constructed, and the construction formula is: ; where d i represents the corresponding eigenvector value, ; N is the original face image, and ; M represents the covariance matrix; T represents the "vectorization" of the image eigenvalues, and all eigenvector values ​​form the eigenface space m; the averaged data is projected in the eigenface space ; Among them, p i is the projection in the eigenface space, i=1, 2, 3...n.

4. The intelligent visual big data access control system according to claim 1 is characterized by: The convolutional neural network of the face extraction module includes a convolution layer, a pooling layer, a dropout layer, a tiling layer, a fully connected layer and a classification layer; The convolution layer receives the face image preprocessed by the face image preprocessing module to obtain input image data , where y1 is the input image, W1 is the width of the input image, H1 is the height of the input image, and D is the number of channels; if the image is RGB type, the number of channels is 3; if the image is a grayscale image, the number of channels is 1; the input image data is obtained after the convolution layer operation to obtain the output image data ; Where y2 is the output image, W2 is the width of the output image, and H2 is the height of the output image; ; ; Where F1 is the width and height of the convolution kernel, S1 is the step size of each convolution operation, and P is the padding number; The pooling layer receives the data y2 after the convolution operation, and outputs the image data after the pooling operation: ,in, , , W3 is the width of the pooling layer output image, H3 is the height of the pooling layer output image, F2 is the width and height of the pooling kernel, and S2 is the step size of each pooling operation; The calculation formula of the Dropout layer is: ; ;Wherein, the Bernoulli function represents the generation probability r vector, i.e., a vector of 0 and 1; Represents the data generated after the input data y3 is dropped; The tiling layer transforms the input data into one dimension. Convert to , the flat layer is the transition layer from convolution, pooling and Dropout to the fully connected layer; The fully connected layer finally obtains the output data , where y5 represents the output data of the fully connected layer, ω is the weight, y4 is the input data, and b represents the offset; The classification layer adopts the softmax classification function, and the formula is expressed as: ,in, represents the output probability, z j represents the input data of each neuron in y5, and J represents the number of all neurons.

5. The intelligent visual big data access control system according to claim 1 is characterized by: The face detection module verifies and identifies the face to be identified, including the following steps: Step S01: Noise filtering is performed on the image of the face to be recognized. The formula is: , where f(x,y) represents the processed face image pixels to be recognized, g(xv,yt) represents the corresponding input face image pixels to be recognized, and W represents 3×3 and 5×5 two-dimensional templates; Step S02: The face to be identified in step S01 is averaged and marked as τ, and the face image to be identified is projected into the feature face space m. The feature vector of the face to be identified is expressed as , where ρ τ is the feature vector of the face to be identified, δ is the average value in the face recognition database; calculate the vector distance between the feature vector of the face to be identified and the feature face space of the face recognition database: ,in, is the vector distance between the feature vector of the face to be identified and the feature face space of the face recognition database; p i is the projection in the eigenface space, i=1, 2, 3...n; n is the number of images of the same face; Step S03: Perform threshold comparison on the vector distance in step S02. If , indicating that the face image to be recognized matches the face image entered into the face recognition database, and the recognition is successful; otherwise, it indicates that the face image to be recognized does not match the face image entered into the face recognition database, and the recognition fails; where YU is the threshold.

6. The intelligent visual big data access control system according to claim 1, characterized in that: The specific method of the background detection module to verify and identify the background face is: Extracting features of background faces , where N is the N dimension, extracting facial features from the face recognition database , where q is the qth face entered; calculate the Euclidean distance between the features of the background face and the features of the face in the face recognition database: , where Dis q is the Euclidean distance between the background face feature and the qth face feature entered into the face recognition database; If there exists a q such that Dis q <UI, it indicates that the background face image matches the qth face image in the face recognition database and the recognition is successful; otherwise, it indicates that the background face image does not match any face in the face recognition database and the recognition fails; the UI is a preset background threshold.

7. The intelligent visual big data access control system according to claim 1, characterized in that: The specific method in which the recognition result processing module analyzes the recognition results of the face detection module and the background detection module is as follows: If the recognition results of the face detection module and the background detection module are both successful, the analysis result is "successful", and the access control release command is sent to the command execution module; If the recognition result of the face detection module is successful, and the recognition result of the background detection module is failed, the analysis result is "there is a hidden danger of unfamiliarity", and an early warning instruction is sent to the early warning instruction module; If the recognition result of the face detection module is failure, no matter whether the recognition result of the background detection module is success or failure, the analysis result is "failure". At this time, the access control release command is not sent to the command execution module, and the access control cannot be released.

Images