A portrait data information collection and analysis system based on big data

Abstract

The present application relates to the technical fields of computer vision and big data processing, in particular to a portrait data information collection and analysis system based on big data, which comprises: a data collection and storage unit using multi-thread data collection technology to collect face images in real time, and storing and indexing the face images; a data image processing unit for pre-processing and standardizing the portrait images collected by the data collection and storage unit; a face feature analysis unit for extracting the features of the standardized face images, analyzing the face image features using spectral graph theory combined with frequency domain analysis algorithm, classifying and updating the index annotation of the face images; and a security management control unit for overall system security and compliance, providing privacy protection, access control and anomaly detection functions. The portrait data information collection and analysis system based on big data realizes portrait data analysis, classification and annotation through the feature fusion technology combining spectral graph theory and frequency domain analysis.

Description

Technical Field

[0001] This invention relates to the field of computer vision and big data processing technology, and more specifically, to a system for collecting and analyzing human facial data based on big data. Background Technology

[0002] The big data-based facial data acquisition and analysis system aims to efficiently manage and retrieve massive amounts of facial data and accurately extract and analyze facial features. By combining distributed index storage with a feature fusion method that integrates spectral theory and frequency domain analysis, it controls the balance between feature extraction accuracy and real-time performance, enabling rapid retrieval, accurate identification, and dynamic updating of index labels for facial information in a multi-source data environment.

[0003] Facial data acquisition and analysis systems often struggle to maintain stable data management and analysis efficiency in high-concurrency, cross-regional data acquisition scenarios. Furthermore, uneven data source distribution, inconsistent image quality, and high complexity of multi-channel feature fusion can lead to insufficient robustness in feature extraction, unstable classification results, and untimely index label updates, thus affecting the accuracy of facial recognition and the efficiency of data retrieval. Therefore, this paper proposes a facial data acquisition and analysis system based on big data. Summary of the Invention

[0004] The purpose of this invention is to provide a human image data information acquisition and analysis system based on big data, in order to solve the problems mentioned in the background art, which are caused by uneven distribution of data sources, inconsistent image quality, and high complexity of multi-channel feature fusion, resulting in insufficient robustness of feature extraction, unstable classification results, and untimely index label updates, thereby affecting the accuracy of human image recognition and the efficiency of data retrieval.

[0005] To achieve the above objectives, the present invention aims to provide a facial data information acquisition and analysis system based on big data, comprising:

[0006] The data acquisition and storage unit uses multi-threaded data acquisition technology to acquire face images in real time, uses the Elasticsearch distributed database to store the face images, and indexes and labels the face images.

[0007] It also includes a data image processing unit, which is used to preprocess and standardize the portrait images acquired by the data acquisition and storage unit;

[0008] It also includes a face feature analysis unit, which is used to extract standardized face image features, analyze face image features using spectral theory combined with frequency domain analysis algorithms, classify and update the index labels of face images;

[0009] It also includes a security management and control unit, which is responsible for the overall security and compliance of the system and provides privacy protection, access control and anomaly detection functions.

[0010] As a further improvement to this technical solution, the data acquisition and storage unit includes a face image acquisition module and a storage and annotation module;

[0011] The face image acquisition module uses a Python multi-threaded programming model to acquire image data from a data source in real time, and uses the MTCNN face detection algorithm to identify and crop the face region to obtain a face image; the data source includes a camera, a mobile device, and recorded video;

[0012] The storage and labeling module stores face images in an Elasticsearch distributed database and generates an index label for each face image. The index label includes time, location, device ID, and person ID.

[0013] As a further improvement to this technical solution, the data image processing unit includes an image quality enhancement module and an image normalization module;

[0014] The image quality enhancement module is used to smooth the face image using Gaussian filtering.

[0015] The image normalization module is used to scale the face image to a predetermined standard size, apply a color normalization algorithm to unify the color distribution of the face image, and generate the final standardized face image.

[0016] As a further improvement to this technical solution, the face feature analysis unit includes a feature extraction module and a feature analysis index update module;

[0017] The feature extraction module uses the Canny edge detection algorithm to extract facial contour features of the face image, the SIFT feature detection algorithm to extract key point distribution features of the face image, and the local binary mode method to extract texture features of the face image; the face image features include facial contour features, key point distribution features, and texture features.

[0018] The feature analysis index update module uses spectral theory combined with frequency domain analysis algorithms to analyze and classify the features of face images and update the index labels of face images.

[0019] As a further improvement to this technical solution, the feature extraction module uses the Canny edge detection algorithm to extract facial contour features of the face image, the SIFT feature detection algorithm to extract key point distribution features of the face image, and the local binary mode method to extract texture features of the face image. The specific steps are as follows:

[0020] S3.1.1. Extract facial contour features from face images using the Canny edge detection algorithm:

[0021] Based on standardized face images Image gradient parallel nonmaximum suppression and double threshold hysteresis processing are calculated to obtain a binary edge map. :

[0022] ;

[0023] in, For face images The pixel coordinates on;

[0024] From binary edge map Extract the maximum connected edge point set That is, the facial contour of a face image; For the first The coordinates of the edge points; This represents the total number of edge points;

[0025] Use an ellipse to fit the contour point set Solve for parameters Facial contour features :

[0026] ;

[0027] in, The x-coordinate of the center point of the ellipse; The ordinate of the center point of the ellipse; The horizontal radius of the ellipse; The perpendicular radius of the ellipse; Let be the rotation angle of the ellipse relative to the horizontal axis;

[0028] S3.1.2. Extract key point distribution features from face images using the SIFT feature detection algorithm:

[0029] Based on standardized face images Construct the Gaussian pyramid and the difference Gaussian pyramid:

[0030] ;

[0031] ;

[0032] in, It is a standard Gaussian filter; For the Gaussian pyramid; For difference Gaussian pyramids; is a constant representing the scale interval; The standard deviation of the Gaussian kernel at the current scale;

[0033] The extreme points in the difference Gaussian pyramid are detected as key points in the face image, resulting in a set of key points. :

[0034] ;

[0035] in, Key point Position coordinates; Key point The scale;

[0036] Calculate key points Gradient directions are distributed in the local neighborhood, and principal directions are assigned. :

[0037] ;

[0038] ;

[0039] ;

[0040] in, Key point A point in the local neighborhood; The direction of the pixel gradient; It is a weighted function; For indicator functions; Key point The main direction;

[0041] At the key point Within the neighborhood, the gradient direction and magnitude statistics form a descriptor vector:

[0042] SIF k =[ h 1 , h 2 ,…, h D s ] ;

[0043] in, Key point Local descriptors; The dimension of the SIFT descriptor; Key point The count value of the neighborhood gradient direction histogram;

[0044] Obtain the key point distribution feature set:

[0045] ;

[0046] in, The total number of key points; This is a set of key point distribution features;

[0047] S3.1.3 Extracting texture features from face images using the local binary pattern method:

[0048] Let the center pixel value be The neighboring pixel value is ;

[0049] Calculate pixels The LBP code is:

[0050] ;

[0051] in, For pixels LBP code; The index of the neighboring pixels; This represents the total number of neighboring pixels. For calculation Indicator functions;

[0052] Calculate the LBP code for all pixels in a face image:

[0053] ;

[0054] in, The LBP code value is The frequency of occurrence of pixels in the entire face image; The value of the LBP code. ; The width of the face image; The height of the face image; For calculation Indicator functions;

[0055] Obtain the LBP histogram vector , This refers to texture features:

[0056] T=[ H LBP (0), H LBP (1),…, H LBP ( 2 N -1)] ;

[0057] in, It is a texture feature.

[0058] As a further improvement to this technical solution, the feature analysis index update module uses spectral theory combined with frequency domain analysis algorithms to analyze and classify the features of face images, and updates the index labels of face images. The specific method steps are as follows:

[0059] S3.2.1, Based on facial contour features and key point distribution feature set We construct a weighted undirected graph using spectral theory.

[0060] S3.2.2 Calculate the degree matrix and adjacency matrix to construct the Laplacian matrix of the graph, and perform eigenvalue decomposition on the Laplacian matrix to extract the spectral eigenvectors of the weighted undirected graph;

[0061] S3.2.3, Texture features of facial images Perform discrete Fourier transform to extract frequency domain features, and then reduce and filter the frequency domain features to obtain the texture frequency domain feature vector;

[0062] S3.2.4. The spectral feature vector and the texture frequency domain feature vector are fused at the feature level into a comprehensive feature vector, and the comprehensive feature vector is classified using a statistical classification method. The index label of the face image is updated according to the classification result.

[0063] As a further improvement to this technical solution, in S3.2.1, based on facial contour features... and key point distribution feature set We construct a weighted undirected graph using spectral graph theory, as follows:

[0064] S3.2.1.1 Treat the key points of the face image as nodes of a graph, and define a node set:

[0065] ;

[0066] in, The first node defined for the first key point. The second node is defined for the second key point, and so on. For the first The first key point defined by One node; A set of nodes constitutes a node set. ;

[0067] S3.2.1.2 Construct a weighted undirected graph based on the Euclidean distance between keypoints and the similarity of SIFT descriptor features:

[0068] Define adjacency matrix :

[0069] ;

[0070] ;

[0071] in, For nodes and nodes The Euclidean distance between them; For nodes SIFT descriptors and nodes The Euclidean distance between SIFT descriptors; To control the distance parameter; To control the feature similarity parameter; Represents a node and nodes The degree of correlation;

[0072] Utilizing facial contour features Obtain the global geometric reference of the face in the image for keypoint location normalization and graph weight calculation; in the calculation First, normalize the coordinates;

[0073] In step S3.2.2, the Laplacian matrix of the graph is constructed by calculating the degree matrix and the adjacency matrix, and eigenvalue decomposition is performed on the Laplacian matrix to extract the spectral eigenvectors of the weighted undirected graph. The specific method is as follows:

[0074] S3.2.2.1 Defining the degree matrix It is a diagonal matrix, degree matrix The diagonal elements are as follows:

[0075] ;

[0076] S3.2.2.2, Define the Laplace matrix:

[0077] ;

[0078] in, It is a Laplace matrix;

[0079] S3.2.2.3, Laplace matrix Perform eigenvalue decomposition:

[0080] ;

[0081] in, Laplace matrix eigenvectors; Laplace matrix eigenvalues;

[0082] eigenvalues Sort by size from smallest to largest, and take the first few. The smaller non-zero eigenvalues ​​and their corresponding eigenvectors form the spectral eigenvectors:

[0083] s=[ λ 1 , λ 2 ,…, λ M ] T ;

[0084] in, These are spectral eigenvectors; This is a transpose operation.

[0085] As a further improvement to this technical solution, in step S3.2.3, the texture features of the face image are... To extract frequency domain features, a discrete Fourier transform is performed. These features are then dimensionality-reduced and filtered to obtain the texture frequency domain feature vector. The specific method is as follows:

[0086] S3.2.3.1 Texture features of facial images The spectrum is obtained by performing a discrete Fourier transform. :

[0087] ;

[0088] F=[F(0),F(1),…,F( 2 N -1)] ;

[0089] in, The LBP code value is The frequency of a pixel appearing in the entire face image; This is the index of the frequency domain components after the discrete Fourier transform; To After performing the Discrete Fourier Transform, at the frequency index The complex spectrum value obtained at the location; The imaginary unit; is the rotation factor of the discrete Fourier transform;

[0090] S3.2.3.2, Before extraction The amplitude of each frequency component is used as the texture frequency domain feature vector. :

[0091] ;

[0092] in, For the spectrum In frequency index The amplitude value at that point; For the spectrum In frequency index The amplitude value at that point; .

[0093] As a further improvement to this technical solution, in S3.2.4, the spectral feature vector and the texture frequency domain feature vector are fused at the feature level into a comprehensive feature vector, and a statistical classification method is used to classify the comprehensive feature vector. The index label of the face image is updated according to the classification result. The specific method is as follows:

[0094] S3.2.4.1, Spectral eigenvectors and texture frequency domain feature vector Perform feature-level fusion to obtain a comprehensive feature vector. :

[0095] c=[s;t]∈ R M+P ;

[0096] in, for One dimension;

[0097] S3.2.4.2, Using Support Vector Machines to synthesize feature vectors Categorize:

[0098] The decision function of a support vector machine is:

[0099] ;

[0100] in, This represents the number of support vectors. For Lagrange multipliers; For kernel function For bias terms; The classification results;

[0101] Classification results This includes the category, identity, and clustering labels of the facial images;

[0102] S3.2.4.3, Based on the classification results The analysis results are then updated in the index annotations of the face images.

[0103] As a further improvement to this technical solution, the security management and control unit includes an access control module, a privacy protection data encryption module, and an anomaly detection and auditing module;

[0104] The access control module is used to perform role-based access control on system users, processes and applications, configure fine-grained permission policies, and implement multi-factor authentication measures.

[0105] The privacy protection data encryption module is used for encrypted storage and transmission of facial image data;

[0106] The anomaly detection and auditing module is used to monitor the system's operating status in real time, continuously audit and analyze access logs, operational behaviors, and data transmission, identify potential security threats and abnormal events using intrusion detection systems and security information and event management tools, and automatically trigger warnings when an anomaly is detected.

[0107] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0108] 1. This big data-based facial data information collection and analysis system, based on a distributed index storage and high-concurrency data processing architecture, can quickly achieve efficient data storage, flexible retrieval, and accurate matching when collecting massive amounts of facial image data from multiple sources and points simultaneously.

[0109] 2. In this big data-based facial data information acquisition and analysis system, the feature fusion technology that combines spectral theory with frequency domain analysis is used to achieve fine classification and annotation updates of facial images. Attached Figure Description

[0110] Figure 1 This is an overall flowchart of the present invention;

[0111] The meanings of the labels in the diagram are as follows:

[0112] 1. Data acquisition and storage unit; 2. Data image processing unit; 3. Facial feature analysis unit; 4. Security management and control unit; 11. Facial image acquisition module; 12. Storage and annotation module; 21. Image quality enhancement module; 22. Image normalization module; 31. Feature extraction module; 32. Feature analysis index update module; 41. Access control module; 42. Privacy protection data encryption module; 43. Anomaly detection and auditing module. Detailed Implementation

[0113] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0114] Example

[0115] Please see Figure 1 As shown, a facial data information collection and analysis system based on big data is provided, including:

[0116] The data acquisition and storage unit 1 uses multi-threaded data acquisition technology to acquire face images in real time, uses the Elasticsearch distributed database to store face images, and indexes and labels the face images.

[0117] In this embodiment, the data acquisition and storage unit 1 includes a face image acquisition module 11 and a storage and annotation module 12;

[0118] The face image acquisition module 11 uses a Python multi-threaded programming model to acquire image data from the data source in real time, and uses the MTCNN face detection algorithm to identify and crop the face region to obtain the face image; the data source includes a camera, a mobile device, and recorded video;

[0119] In this embodiment, the face image acquisition module 11 uses a Python multithreaded programming model to acquire face images from the data source in real time. The specific steps are as follows:

[0120] Establish stable data stream connections by accessing various cameras and mobile devices via RTSP, HTTP, or other protocols; allocate independent acquisition threads for each data source to receive video streams or image frames in real time; extract continuous image frames from the video stream; and temporarily store the acquired image frames using a Python thread-safe queue (Queue).

[0121] In this embodiment, the MTCNN face detection algorithm is a face detection and alignment algorithm that uses a cascaded multi-task convolutional neural network to perform face detection and key point localization simultaneously.

[0122] The storage labeling module 12 stores the face images in the Elasticsearch distributed database and generates an index label for each face image. The index label includes time, location, device ID, and person ID.

[0123] In this embodiment, multiple Elasticsearch nodes are deployed to build a distributed database cluster, and a multi-field index based on time, location, and person ID is designed and created. Face images are stored in Base64 encoding, and face image data and its metadata are inserted in batches through the Elasticsearch API.

[0124] It also includes a data image processing unit 2, which is used to preprocess and standardize the portrait images acquired by the data acquisition and storage unit 1;

[0125] In this embodiment, the data image processing unit 2 includes an image quality enhancement module 21 and an image normalization module 22;

[0126] The image quality enhancement module 21 is used to smooth the face image by using Gaussian filtering;

[0127] The image normalization module 22 is used to scale the face image to a predetermined standard size, apply a color normalization algorithm to unify the color distribution of the face image, and generate the final standardized face image.

[0128] In this embodiment, the predetermined standard size is 224x224 pixels, and the image background is removed or replaced according to application requirements to highlight the main face and reduce the interference of the background on feature extraction.

[0129] It also includes a face feature analysis unit 3, which is used to extract standardized face image features, analyze face image features using spectral theory combined with frequency domain analysis algorithm, classify and update the index label of face image;

[0130] In this embodiment, the face feature analysis unit 3 includes a feature extraction module 31 and a feature analysis index update module 32;

[0131] The feature extraction module 31 uses the Canny edge detection algorithm to extract facial contour features of the face image, the SIFT feature detection algorithm to extract key point distribution features of the face image, and the local binary mode method to extract texture features of the face image; the face image features include facial contour features, key point distribution features, and texture features.

[0132] The feature analysis index update module 32 uses spectral theory combined with frequency domain analysis algorithm to analyze and classify the features of face images and update the index labels of face images.

[0133] The feature extraction module 31 uses the Canny edge detection algorithm to extract facial contour features of the face image, the SIFT feature detection algorithm to extract key point distribution features of the face image, and the local binary mode method to extract texture features of the face image. The specific steps are as follows:

[0134] S3.1.1. Extract facial contour features from face images using the Canny edge detection algorithm:

[0135] Based on standardized face images Image gradient parallel nonmaximum suppression and double threshold hysteresis processing are calculated to obtain a binary edge map. :

[0136] ;

[0137] in, For face images The pixel coordinates on;

[0138] From binary edge map Extract the maximum connected edge point set That is, the facial contour of a face image; For the first The coordinates of the edge points; This represents the total number of edge points;

[0139] Use an ellipse to fit the contour point set Solve for parameters Facial contour features :

[0140] ;

[0141] in, The x-coordinate of the center point of the ellipse; The ordinate of the center point of the ellipse; The horizontal radius of the ellipse; The perpendicular radius of the ellipse; Let be the rotation angle of the ellipse relative to the horizontal axis;

[0142] S3.1.2. Extract key point distribution features from face images using the SIFT feature detection algorithm:

[0143] Based on standardized face images Construct the Gaussian pyramid and the difference Gaussian pyramid:

[0144] ;

[0145] ;

[0146] in, It is a standard Gaussian filter; For the Gaussian pyramid; For difference Gaussian pyramids; is a constant representing the scale interval; The standard deviation of the Gaussian kernel at the current scale;

[0147] The extreme points in the difference Gaussian pyramid are detected as key points in the face image, resulting in a set of key points. :

[0148] ;

[0149] in, Key point Position coordinates; Key point The scale;

[0150] Calculate key points Gradient directions are distributed in the local neighborhood, and principal directions are assigned. :

[0151] ;

[0152] ;

[0153] ;

[0154] in, Key point A point in the local neighborhood; The direction of the pixel gradient; It is a weighted function; For indicator functions; Key point The main direction;

[0155] In this embodiment, For calculation The indicator function, when The function returns 1 if the condition is met; otherwise, it returns 0. Used for statistical analysis of gradient direction in images The number of pixels;

[0156] At the key point Within the neighborhood, the gradient direction and magnitude statistics form a descriptor vector:

[0157] SIF k =[ h 1 , h 2 ,…, h D s ] ;

[0158] in, Key point Local descriptors; The dimension of the SIFT descriptor; Key point The count value of the neighborhood gradient direction histogram;

[0159] Obtain the key point distribution feature set:

[0160] ;

[0161] in, The total number of key points; This is a set of key point distribution features;

[0162] S3.1.3 Extracting texture features from face images using the local binary pattern method:

[0163] Let the center pixel value be The neighboring pixel value is ;

[0164] Calculate pixels The LBP code is:

[0165] ;

[0166] in, For pixels LBP code; The index of the neighboring pixels; This represents the total number of neighboring pixels. For calculation Indicator functions;

[0167] Calculate the LBP code for all pixels in a face image:

[0168] ;

[0169] in, The LBP code value is The frequency of occurrence of pixels in the entire face image; The value of the LBP code. ; The width of the face image; The height of the face image; For calculation Indicator functions;

[0170] Obtain the LBP histogram vector , This refers to texture features:

[0171] T=[ H LBP (0), H LBP (1),…, H LBP ( 2 N -1)] ;

[0172] in, It is a texture feature.

[0173] The feature analysis index update module 32 uses spectral theory combined with frequency domain analysis algorithms to analyze and classify the features of the face image, and updates the index label of the face image. The specific steps are as follows:

[0174] S3.2.1, Based on facial contour features and key point distribution feature set We construct a weighted undirected graph using spectral theory.

[0175] S3.2.2 Calculate the degree matrix and adjacency matrix to construct the Laplacian matrix of the graph, and perform eigenvalue decomposition on the Laplacian matrix to extract the spectral eigenvectors of the weighted undirected graph;

[0176] S3.2.3, Texture features of facial images Perform discrete Fourier transform to extract frequency domain features, and then reduce and filter the frequency domain features to obtain the texture frequency domain feature vector;

[0177] S3.2.4. The spectral feature vector and the texture frequency domain feature vector are fused at the feature level into a comprehensive feature vector, and the comprehensive feature vector is classified using a statistical classification method. The index label of the face image is updated according to the classification result.

[0178] In this embodiment S3.2.1, based on facial contour features and key point distribution feature set We construct a weighted undirected graph using spectral graph theory, as follows:

[0179] S3.2.1.1 Treat the key points of the face image as nodes of a graph, and define a node set:

[0180] ;

[0181] in, The first node defined for the first key point. The second node is defined for the second key point, and so on. For the first The first key point defined by One node; A set of nodes constitutes a node set. ;

[0182] S3.2.1.2 Construct a weighted undirected graph based on the Euclidean distance between keypoints and the similarity of SIFT descriptor features:

[0183] Define adjacency matrix :

[0184] ;

[0185] ;

[0186] in, For nodes and nodes The Euclidean distance between them; For nodes SIFT descriptors and nodes The Euclidean distance between SIFT descriptors; To control the distance parameter; To control the feature similarity parameter; Represents a node and nodes The degree of correlation;

[0187] Utilizing facial contour features Obtain the global geometric reference of the face in the image for keypoint location normalization and graph weight calculation; in the calculation First, normalize the coordinates;

[0188] In this embodiment S3.2.2, the degree matrix and adjacency matrix are used to construct the Laplacian matrix of the graph, and eigenvalue decomposition is performed on the Laplacian matrix to extract the spectral eigenvectors of the weighted undirected graph. The specific method is as follows:

[0189] S3.2.2.1 Defining the degree matrix It is a diagonal matrix, degree matrix The diagonal elements are as follows:

[0190] ;

[0191] S3.2.2.2, Define the Laplace matrix:

[0192] ;

[0193] in, It is a Laplace matrix;

[0194] S3.2.2.3, Laplace matrix Perform eigenvalue decomposition:

[0195] ;

[0196] in, Laplace matrix eigenvectors; Laplace matrix eigenvalues;

[0197] eigenvalues Sort by size from smallest to largest, and take the first few. The smaller non-zero eigenvalues ​​and their corresponding eigenvectors form the spectral eigenvectors:

[0198] s=[ λ 1 , λ 2 ,…, λ M ] T ;

[0199] in, These are spectral eigenvectors; This is a transpose operation.

[0200] In this embodiment S3.2.3, the texture features of the face image are... To extract frequency domain features, a discrete Fourier transform is performed. These features are then dimensionality-reduced and filtered to obtain the texture frequency domain feature vector. The specific method is as follows:

[0201] S3.2.3.1 Texture features of facial images The spectrum is obtained by performing a discrete Fourier transform. :

[0202] ;

[0203] F=[F(0),F(1),…,F( 2 N -1)] ;

[0204] in, The LBP code value is The frequency of a pixel appearing in the entire face image; This is the index of the frequency domain components after the discrete Fourier transform; To After performing the Discrete Fourier Transform, at the frequency index The complex spectrum value obtained at the location; The imaginary unit; is the rotation factor of the discrete Fourier transform;

[0205] S3.2.3.2, Before extraction The amplitude of each frequency component is used as the texture frequency domain feature vector. :

[0206] ;

[0207] in, For the spectrum In frequency index The amplitude value at that point; For the spectrum In frequency index The amplitude value at that point; .

[0208] In this embodiment S3.2.4, the spectral feature vector and the texture frequency domain feature vector are fused at the feature level into a comprehensive feature vector, and a statistical classification method is used to classify the comprehensive feature vector. The index label of the face image is updated according to the classification result. The specific method is as follows:

[0209] S3.2.4.1, Spectral eigenvectors and texture frequency domain feature vector Perform feature-level fusion to obtain a comprehensive feature vector. :

[0210] c=[s;t]∈ R M+P ;

[0211] in, for One dimension;

[0212] S3.2.4.2, Using Support Vector Machines to synthesize feature vectors Categorize:

[0213] The decision function of a support vector machine is:

[0214] ;

[0215] in, This represents the number of support vectors. For Lagrange multipliers; For kernel function For bias terms; The classification results;

[0216] Classification results This includes the category, identity, and clustering labels of the facial images;

[0217] S3.2.4.3, Based on the classification results The analysis results are then updated in the index annotations of the face images.

[0218] It also includes a security management and control unit 4, which is responsible for the overall security and compliance of the system and provides privacy protection, access control and anomaly detection functions;

[0219] In this embodiment, the security management and control unit 4 includes an access control module 41, a privacy protection data encryption module 42, and an anomaly detection and auditing module 43;

[0220] The access control module 41 is used to perform role-based access control on system users, processes and applications, configure fine-grained permission policies, and implement multi-factor authentication measures.

[0221] In this embodiment, the fine-grained permission policy is designed as follows:

[0222] System access is divided into multiple levels of roles and resource types, and the scope of permissions for each role to the resources is clearly defined:

[0223] Character type:

[0224] Administrator: Has the highest level of privileges, and can manage user accounts, assign roles, modify security policies, and configure data sources and storage policies;

[0225] Data maintainer: Responsible for data cleaning, labeling and preprocessing, has write and edit permissions for data processing-related operations, but does not have the right to modify security policies or access sensitive data in plaintext;

[0226] Analysts: Primarily perform read-only or restricted write operations such as data querying, retrieval, feature analysis, and model training;

[0227] Read-only users can only query and retrieve specific datasets; they cannot modify or download the original image data, nor can they access sensitive metadata.

[0228] Resource type:

[0229] Data resources: include raw face images, feature data, and indexed and labeled metadata;

[0230] System configuration resources include permission policy files, security configuration files, logs, and audit records.

[0231] Management operations include adding users, modifying permission policies, and creating or deleting indexes.

[0232] The multi-factor authentication measures are designed as follows:

[0233] Multi-factor authentication requires users to provide at least two different types of identity elements when logging in, escalating privileges, or performing sensitive operations:

[0234] Knowledge factors include passwords and PIN codes;

[0235] Possession factors include one-time dynamic verification codes for mobile phones, USB tokens, or security keys;

[0236] Biometrics include fingerprints, facial recognition, or voiceprint verification, but must comply with privacy regulations;

[0237] Initial login and periodic verification: When a user logs in for the first time and accesses sensitive datasets again at fixed intervals, they need to enter their username and password and obtain a one-time verification code via mobile app or SMS, or use a dynamic password generated by hardware token for secondary verification before they can enter the system.

[0238] Two-factor authentication during permission upgrade: When a regular analyst attempts to perform advanced analysis operations, the system requires them to provide an additional one-time token or software / hardware certificate signature.

[0239] Geographic location and device fingerprinting: Optional geographic restrictions and device fingerprinting technology can be introduced; when a user attempts to access data in an unusual location or on an unidentified device, the system will trigger the MFA process and may require administrator approval.

[0240] Expiration and Re-authentication: To prevent sessions from remaining valid for a long time, a second factor must be entered after a certain period of time to ensure that operations are only performed in active and trusted sessions;

[0241] The privacy protection data encryption module 42 is used to encrypt and store and transmit facial image data.

[0242] The anomaly detection and auditing module 43 is used to monitor the system's operating status in real time, continuously audit and analyze access logs, operational behaviors, and data transmission, identify potential security threats and abnormal events using intrusion detection systems and security information and event management tools, and automatically trigger warnings when an anomaly is detected.

[0243] In this embodiment, an intrusion detection system is a security tool that monitors network or system activity and alerts to events that violate security policies or display suspicious behavior; a security information and event management tool is a system that integrates, analyzes, and correlates security events and log data from multiple sources.

[0244] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A facial data information collection and analysis system based on big data, characterized in that, include: The data acquisition and storage unit (1) uses multi-threaded data acquisition technology to acquire face images in real time, uses Elasticsearch distributed database to store face images, and indexes and labels the face images. The data image processing unit (2) is used to preprocess and standardize the portrait images acquired by the data acquisition and storage unit (1); The face feature analysis unit (3) is used to extract standardized face image features, analyze face image features using spectral theory combined with frequency domain analysis algorithm, classify and update the index label of face image; Security management and control unit (4), which is responsible for the overall security and compliance of the system and provides privacy protection, access control and anomaly detection functions; The face feature analysis unit (3) includes a feature extraction module (31) and a feature analysis index update module (32). The feature analysis index update module (32) uses spectral theory combined with frequency domain analysis algorithm to analyze and classify the features of face images and update the index labels of face images. The specific steps are as follows: S3.2.1, Based on facial contour features and key point distribution feature set We construct a weighted undirected graph using spectral theory. S3.2.2 Calculate the degree matrix and adjacency matrix to construct the Laplacian matrix of the graph, and perform eigenvalue decomposition on the Laplacian matrix to extract the spectral eigenvectors of the weighted undirected graph; S3.2.3, Texture features of facial images Perform discrete Fourier transform to extract frequency domain features, and then reduce and filter the frequency domain features to obtain the texture frequency domain feature vector; S3.2.

4. The spectral feature vector and the texture frequency domain feature vector are fused at the feature level into a comprehensive feature vector, and the comprehensive feature vector is classified using a statistical classification method. The index label of the face image is updated according to the classification result.

2. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: The data acquisition and storage unit (1) includes a face image acquisition module (11) and a storage and annotation module (12). The face image acquisition module (11) uses a Python multi-threaded programming model to acquire image data from the data source in real time, and uses the MTCNN face detection algorithm to identify and crop the face region to obtain the face image; the data source includes cameras, mobile devices and recorded videos; The storage labeling module (12) stores the face images in the Elasticsearch distributed database and generates index labels for each face image. The index labels include time, location, device ID and person ID.

3. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: The data image processing unit (2) includes an image quality enhancement module (21) and an image normalization module (22). The image quality enhancement module (21) is used to smooth the face image by Gaussian filtering; The image normalization module (22) is used to scale the face image to a predetermined standard size, apply a color normalization algorithm to unify the color distribution of the face image, and generate the final standardized face image.

4. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: The feature extraction module (31) uses the Canny edge detection algorithm to extract facial contour features of the face image, uses the SIFT feature detection algorithm to extract key point distribution features of the face image, and uses the local binary mode method to extract texture features of the face image; the face image features include facial contour features, key point distribution features, and texture features; The feature analysis index update module (32) uses spectral theory combined with frequency domain analysis algorithm to analyze and classify the features of face images and update the index label of face images.

5. The facial data information acquisition and analysis system based on big data according to claim 4, characterized in that: The feature extraction module (31) uses the Canny edge detection algorithm to extract facial contour features of the face image, uses the SIFT feature detection algorithm to extract key point distribution features of the face image, and uses the local binary mode method to extract texture features of the face image. The specific steps are as follows: S3.1.

1. Extract facial contour features from face images using the Canny edge detection algorithm: Based on standardized face images Image gradient parallel nonmaximum suppression and double threshold hysteresis processing are calculated to obtain a binary edge map. : ； in, For face images The pixel coordinates on; From binary edge map Extract the maximum connected edge point set That is, the facial contour of a face image; For the first The coordinates of the edge points; This represents the total number of edge points; Use an ellipse to fit the contour point set Solve for parameters Facial contour features : ； in, The x-coordinate of the center point of the ellipse; The ordinate of the center point of the ellipse; The horizontal radius of the ellipse; The perpendicular radius of the ellipse; Let be the rotation angle of the ellipse relative to the horizontal axis; S3.1.

2. Extract key point distribution features from face images using the SIFT feature detection algorithm: Based on standardized face images Construct the Gaussian pyramid and the difference Gaussian pyramid: ； ； in, It is a standard Gaussian filter; For the Gaussian Pyramid; For difference Gaussian pyramids; is a constant representing the scale interval; The standard deviation of the Gaussian kernel at the current scale; The extreme points in the difference Gaussian pyramid are detected as key points in the face image, resulting in a set of key points. : ； in, Key point Position coordinates; Key point The scale; Calculate key points Gradient directions are distributed in the local neighborhood, and principal directions are assigned. : ； ； ； in, Key point A point in the local neighborhood; The direction of the pixel gradient; It is a weighted function; For indicator functions; Key point The main direction; At the key point Within the neighborhood, the gradient direction and magnitude statistics form a descriptor vector: ； in, Key point Local descriptors; The dimension of the SIFT descriptor; Key point The count value of the neighborhood gradient direction histogram; Obtain the key point distribution feature set: ； in, The total number of key points; This is a set of key point distribution features; S3.1.3 Extracting texture features from face images using the local binary pattern method: Let the center pixel value be The neighboring pixel value is ; Calculate pixels The LBP code is: ； in, For pixels LBP code; The index of the neighboring pixels; This represents the total number of neighboring pixels. For calculation Indicator functions; Calculate the LBP code for all pixels in a face image: ； in, The LBP code value is The frequency of occurrence of pixels in the entire face image; The value of the LBP code. ; The width of the face image; The height of the face image; For calculation Indicator functions; Obtain the LBP histogram vector , This refers to texture features: ； in, It is a texture feature.

6. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: In S3.2.1, based on facial contour features and key point distribution feature set We construct a weighted undirected graph using spectral graph theory, as follows: S3.2.1.1 Treat the key points of the face image as nodes of a graph, and define a node set. : ； in, The first node defined for the first key point. The second node is defined for the second key point, and so on. For the first The first key point defined by One node; A set of nodes constitutes a node set. ; S3.2.1.2 Construct a weighted undirected graph based on the Euclidean distance between keypoints and the similarity of SIFT descriptor features: Define adjacency matrix : ； ； in, For nodes and nodes The Euclidean distance between them; For nodes SIFT descriptors and nodes The Euclidean distance between SIFT descriptors; To control the distance parameter; To control the feature similarity parameter; Represents a node and nodes The degree of correlation; Utilizing facial contour features Obtain the global geometric reference of the face in the image for keypoint location normalization and graph weight calculation; in the calculation First, normalize the coordinates; In step S3.2.2, the Laplacian matrix of the graph is constructed by calculating the degree matrix and the adjacency matrix, and eigenvalue decomposition is performed on the Laplacian matrix to extract the spectral eigenvectors of the weighted undirected graph. The specific method is as follows: S3.2.2.1 Defining the degree matrix It is a diagonal matrix, degree matrix The diagonal elements are as follows: ； S3.2.2.2, Define the Laplace matrix: ； in, It is a Laplace matrix; S3.2.2.3, Laplace matrix Perform eigenvalue decomposition: ； in, Laplace matrix eigenvectors; Laplace matrix eigenvalues; eigenvalues Sort by size from smallest to largest, and take the first few. The smaller non-zero eigenvalues ​​and their corresponding eigenvectors form the spectral eigenvectors: ； in, These are spectral eigenvectors; This is a transpose operation.

7. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: In step S3.2.3, the texture features of the face image are... To extract frequency domain features, a discrete Fourier transform is performed. These features are then dimensionality-reduced and filtered to obtain the texture frequency domain feature vector. The specific method is as follows: S3.2.3.1 Texture features of facial images The spectrum is obtained by performing a discrete Fourier transform. : ； ； in, The LBP code value is The frequency of a pixel appearing in the entire face image; This is the index of the frequency domain components after the discrete Fourier transform; To After performing the Discrete Fourier Transform, at the frequency index The complex spectrum value obtained at the location; The imaginary unit; The rotation factor of the discrete Fourier transform; S3.2.3.2, Before extraction The amplitude of each frequency component is used as the texture frequency domain feature vector. : ； in, For the spectrum In frequency index The amplitude value at that point; and so on, For the spectrum In frequency index The amplitude value at that point; .

8. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: In step S3.2.4, the spectral feature vector and the texture frequency domain feature vector are fused at the feature level into a comprehensive feature vector. A statistical classification method is then used to classify the comprehensive feature vector, and the index label of the face image is updated based on the classification result. The specific method is as follows: S3.2.4.1, Spectral eigenvectors and texture frequency domain feature vector Perform feature-level fusion to obtain a comprehensive feature vector. : ； in, for One dimension; S3.2.4.2, Using Support Vector Machines to synthesize feature vectors Categorize: The decision function of a support vector machine is: ； in, This represents the number of support vectors. For Lagrange multipliers; For kernel functions; For bias terms; The classification results; Classification results This includes the category, identity, and clustering labels of the facial images; S3.2.4.3, Based on the classification results The analysis results are then updated in the index annotations of the face images.

9. The facial data information acquisition and analysis system based on big data according to claim 1, characterized in that: The security management and control unit (4) includes an access control module (41), a privacy protection data encryption module (42), and an anomaly detection and auditing module (43). The access control module (41) is used to perform role-based access control on system users, processes and applications, configure fine-grained permission policies, and implement multi-factor authentication measures. The privacy protection data encryption module (42) is used to encrypt and store and transmit facial image data; The anomaly detection and auditing module (43) is used to monitor the system's operating status in real time, continuously audit and analyze access logs, operational behaviors and data transmission, identify potential security threats and abnormal events using intrusion detection systems and security information and event management tools, and automatically trigger warnings when an anomaly is detected.

Images