CN117994835A - Security face recognition method and device based on deep learning - Google Patents
Security face recognition method and device based on deep learning Download PDFInfo
- Publication number
- CN117994835A CN117994835A CN202410248041.6A CN202410248041A CN117994835A CN 117994835 A CN117994835 A CN 117994835A CN 202410248041 A CN202410248041 A CN 202410248041A CN 117994835 A CN117994835 A CN 117994835A
- Authority
- CN
- China
- Prior art keywords
- face image
- face recognition
- security
- layer
- real
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013135 deep learning Methods 0.000 title claims abstract description 34
- 210000000697 sensory organ Anatomy 0.000 claims abstract description 19
- 239000013598 vector Substances 0.000 claims description 39
- 230000006870 function Effects 0.000 claims description 15
- 238000013527 convolutional neural network Methods 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 210000001508 eye Anatomy 0.000 claims description 8
- 210000004709 eyebrow Anatomy 0.000 claims description 5
- 238000005286 illumination Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 210000000214 mouth Anatomy 0.000 claims description 4
- 210000001331 nose Anatomy 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 2
- 238000011161 development Methods 0.000 description 7
- 230000018109 developmental process Effects 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 210000000887 face Anatomy 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 108010001267 Protein Subunits Proteins 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000003679 aging effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The application relates to a security face recognition method and a security face recognition device based on deep learning, wherein the security face recognition method recognizes a face through a trained security face recognition model, the security face recognition model comprises a generator and a classifier, and the security face recognition method comprises the following steps: acquiring a real face image to be identified and configuration information thereof, wherein the configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image; inputting the real face image and configuration information thereof into a generator, and obtaining a generated face image corresponding to the real face image through the generator; and inputting the generated face image into a classifier, and obtaining a recognition classification result of the real face image through the classifier. By means of configuration information, the influence of the pose and the expression of the face in the real face image on face recognition can be relieved, face recognition accuracy is improved, and the problem that the existing security face recognition method based on deep learning is low in face recognition accuracy is solved.
Description
Technical Field
The application relates to the field of image recognition, in particular to a security face recognition method and device based on deep learning.
Background
In recent years, with the development of security video technology and the rising of artificial intelligence, people have increasingly demanded intelligent life, especially safe life, which is no longer satisfied with the traditional living environment, and increasingly pay attention to personal safety and property safety, and have put higher demands on the safety aspect of public places. Meanwhile, along with the rapid development of economy, more and more population flows into cities, so that the population quantity in the cities is increased dramatically, and how to ensure the public security in the cities is stable becomes a problem to be solved urgently. In order to ensure safety and prevent adverse events, intelligent security has become a current development trend, and video monitoring is an important step of intelligent security, so that a large amount of manpower, equipment and fund investment can be saved, and continuous and omnibearing monitoring service is provided for social life every day. The video is monitored in real time, records can be called at any time, illegal activities are conveniently reported, behaviors of staff in an activity area are restrained, a deterrent effect is exerted, key people are identified in advance, and countermeasures are formulated in advance. Identification is an important component of intelligent security, face recognition is a key technology of identification, face recognition is a highly non-invasive biological identification, identity authentication is performed by analyzing and comparing face visual characteristic information, static images or surveillance videos are input in a query scene, and one or more identities in the scene are identified or verified by using an entered face database. Face recognition is rapidly becoming an important component of entertainment in the fields of video monitoring, such as border control, suspicious tracking and recognition, as well as security aspects of system login, banking, file encryption and the like, man-machine interaction, 3D animation, virtual reality and the like. The development of deep learning brings revolutionary performance improvement to various computer vision tasks, wherein the face recognition based on CNN exceeds human performance, and along with the development of three-dimensional face model technology, the security monitoring face recognition can gradually process various changing factors in the face generation process, and has certain research significance and good development prospect.
Although many experts and scholars have conducted a great deal of research on security face recognition methods based on deep learning at present, and the security face recognition methods aiming at deep learning have broken through in intelligent monitoring environments, existing solutions have many defects, such as posture change, facial expression, aging effect and natural shielding caused by different illumination of people, and the existing deep neural network can lose some similarity information of face recognition through embedding of network learning, so that the accuracy of security face recognition is greatly reduced.
Aiming at the problem of low face recognition accuracy of the existing security face recognition method based on deep learning, no effective solution is proposed at present.
Disclosure of Invention
The invention provides a security face recognition method and device based on deep learning, which are used for solving the problem that the existing security face recognition method based on deep learning is low in face recognition accuracy.
In a first aspect, the present invention provides a security face recognition method based on deep learning, the face is recognized by a trained security face recognition model, the security face recognition model includes a generator and a classifier, and the security face recognition method includes:
Acquiring a real face image to be identified and configuration information thereof, wherein the configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image;
inputting the real face image and configuration information thereof into the generator, and obtaining a generated face image corresponding to the real face image through the generator;
and inputting the generated face image into the classifier, and obtaining a recognition classification result of the real face image through the classifier.
In some embodiments, the acquiring the real face image to be identified and the configuration information thereof includes:
acquiring a real face image to be identified;
normalizing the illumination variation in the real face image;
and adjusting the normalized pixel intensity value of the real face image by adopting gamma correction.
In some embodiments, the acquiring the real face image to be identified and the configuration information thereof further includes:
and positioning the five sense organs in the real face image by adopting a multi-task cascade convolutional neural network, and calculating the distance characteristics between any two parts in the five sense organs to obtain the configuration information of the real face image, wherein the five sense organs comprise eyebrows, eyes, nose, mouth and chin.
In some of these embodiments, the generator comprises a feature extractor, a first encoder, a second encoder, and a decoder;
Inputting the real face image and the configuration information thereof into the generator, obtaining a generated face image corresponding to the real face image through the generator, and comprising the following steps:
Inputting the real face image into the feature extractor, taking the output of the feature extractor as the input of the first encoder, and inputting the configuration information into the second encoder;
Taking the output of the first encoder and the output of the second encoder as the input of a mean value layer and a variance layer, obtaining a mean value vector through the mean value layer, and obtaining a variance vector through the variance layer;
Obtaining potential vectors according to the mean vector and the variance vector;
and inputting the potential vector into the decoder, and obtaining the generated face image through the decoder.
In some of these embodiments, the feature extractor consists of three parallel convolutional neural networks without weight sharing;
The filter sizes of the three convolutional neural networks are 11×11, 7×7 and 5×5 respectively, each convolutional neural network comprises three layers of first sub-networks, and each layer of first sub-network comprises a convolutional layer, a maximum pooling layer and a ReLU activation function which are sequentially connected.
In some of these embodiments, the first encoder and the second encoder each comprise a three-layer second subnetwork;
the third layer of the second sub-network is respectively provided with 4096, 2048 and 1024 neurons, and the third layer of the second sub-network is connected with the mean layer and the variance layer;
the mean layer and the variance layer each have 64 neurons.
In some of these embodiments, the decoder includes a third sub-network of three layers;
the third subnetwork of three layers has 1024, 2048 and 4096 neurons, respectively.
In some embodiments, the classifier includes a hidden layer, a hyperbolic tangent activation function, and an output layer for class prediction connected in sequence;
The output layer uses Softmax activation to represent the confidence level of each class prediction.
In some embodiments, the loss function of the security face recognition model is:
where z represents the potential vector, x represents the data point, μ represents the mean vector, ε represents the variance vector.
In a second aspect, the present invention provides a security face recognition device based on deep learning, which recognizes a face through a trained security face recognition model, where the security face recognition model includes a generator and a classifier, and the security face recognition device includes:
The data acquisition module is used for acquiring a real face image to be identified and configuration information thereof, wherein the configuration information comprises distance features between any two parts in the five sense organs of the real face image;
The image generation module is used for inputting the real face image and the configuration information thereof into the generator, and obtaining a generated face image corresponding to the real face image through the generator;
and the image classification module is used for inputting the generated face image into the classifier, and obtaining the recognition classification result of the real face image through the classifier.
Compared with the related art, the security face recognition method and device based on the deep learning provided by the invention mainly carry out face recognition on the real face image through the security face recognition model based on the deep learning. In addition to taking the real face image as the input of the security face recognition model, the configuration information corresponding to the real face image is also taken as the input of the security face recognition model. The configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image. By the configuration information, the influence of the pose and the expression of the face in the real face image on the face recognition can be relieved, so that the face recognition accuracy is improved, and the problem of low face recognition accuracy in the existing security face recognition method based on deep learning is solved.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the other features, objects, and advantages of the application.
Drawings
Fig. 1 is a flowchart of a security face recognition method based on deep learning provided in an embodiment of the present invention;
FIG. 2 is a specific flow chart of step S110 in some embodiments of the invention;
FIG. 3 is a specific flow chart of step S120 in some embodiments of the invention;
Fig. 4 is a block diagram of a security face recognition device based on deep learning according to an embodiment of the present invention.
Detailed Description
The present application will be described and illustrated with reference to the accompanying drawings and examples for a clearer understanding of the objects, technical solutions and advantages of the present application.
Unless defined otherwise, technical or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms "a," "an," "the," "these" and similar terms in this application are not intended to be limiting in number, but may be singular or plural. The terms "comprising," "including," "having," and any variations thereof, as used herein, are intended to encompass non-exclusive inclusion; for example, a process, method, and system, article, or apparatus that comprises a list of steps or modules (units) is not limited to the list of steps or modules (units), but may include other steps or modules (units) not listed or inherent to such process, method, article, or apparatus. The terms "connected," "coupled," and the like in this disclosure are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as used herein means two or more. "and/or" describes an association relationship of an association object, meaning that there may be three relationships, e.g., "a and/or B" may mean: a exists alone, A and B exist together, and B exists alone. Typically, the character "/" indicates that the associated object is an "or" relationship. The terms "first," "second," "third," and the like, as referred to in this disclosure, merely distinguish similar objects and do not represent a particular ordering for objects.
The embodiment of the invention provides a security face recognition method based on deep learning, which is used for recognizing a face through a trained security face recognition model, wherein the security face recognition model comprises a generator and a classifier. Fig. 1 is a flowchart of a security face recognition method based on deep learning, provided in an embodiment of the present invention, as shown in fig. 1, the flowchart includes the following steps:
Step S110, a real face image to be recognized and configuration information thereof are obtained, wherein the configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image.
Step S120, inputting the real face image and the configuration information thereof into a generator, and obtaining a generated face image corresponding to the real face image through the generator.
Step S130, inputting the generated face image into a classifier, and obtaining a recognition classification result of the real face image through the classifier.
In the technical scheme, the face recognition is carried out on the real face image mainly through the security face recognition model based on deep learning. In addition to taking the real face image as the input of the security face recognition model, the configuration information corresponding to the real face image is also taken as the input of the security face recognition model. The configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image. For example, the distance feature d=102 between the eyebrow and the chin. This allows the network to maximize the pairwise correlation between the front and contour faces and project the data to a common embedding. By the configuration information, the influence of the pose and the expression of the face in the real face image on the face recognition can be relieved, so that the face recognition accuracy is improved, and the problem of low face recognition accuracy in the existing security face recognition method based on deep learning is solved.
Fig. 2 is a specific flowchart of step S110 in some embodiments of the invention. Referring to fig. 2, in some embodiments thereof, step S110, acquiring a real face image to be identified and configuration information thereof includes:
Step S111, acquiring a real face image to be recognized.
Step S112, the illumination change in the real face image is normalized.
Step S113, adopting gamma correction to adjust the pixel intensity value of the normalized real face image.
In this embodiment, after the real face image is acquired, it is also preprocessed, so that the image quality is improved. Because the acquired initial image is large in size and possibly contains a plurality of face areas, different face areas in the initial image can be further segmented, and corresponding real face images are obtained.
Illustratively, the face region in the initial image may be segmented into 128 x 128 pixel real face images, and then the illumination changes in the real face images normalized to reduce the effects of light conditions of the scene, such as shadows, overexposure or underexposure, and then gamma correction is used to adjust the pixel intensity values.
Through the image preprocessing, the quality of the real face image can be improved, so that the face recognition accuracy of the security face recognition model on the real face image is further improved.
Further, in some embodiments, step S110, acquiring the real face image to be identified and the configuration information thereof, further includes:
Step S114, a multi-task cascade convolutional neural network is adopted to locate the five sense organs in the real face image, and the distance characteristics between any two parts in the five sense organs are calculated to obtain the configuration information of the real face image, wherein the five sense organs comprise eyebrows, eyes, noses, mouths and chin.
Specifically, any two parts may be eyes and eyebrows, eyes and nose, eyes and mouth, and the like. Illustratively, a straight line can be drawn using a straight line equation using two points (x 1,y1) and (x 2,y2) of the eyes in the real face image, and a specific formula can be expressed as:
Ax+By+C＝0
(y2-y1)x+(x2-x1)y+(x1y2-x2y1)＝0
Wherein a= (y 2-y1),B＝(x2-x1),C＝(x1y2-x2y1), the third point (x 3,y3) is orthogonal to the straight line formed by the two points of the eye (x 1,y1) and (x 2,y2), and can be expressed as:
In the process of calculating the distance characteristics, grid subdivision and triangle interpolation can be carried out on the real face image, the regularity of the grid is checked by calculating the connection number of each vertex, and the vertices and the connections are added when subdivision iterates are carried out each time.
By conducting grid subdivision and triangle interpolation on the real face image, data enhancement can be achieved, and face recognition accuracy of the security face recognition model is greatly improved.
Fig. 3 is a specific flowchart of step S120 in some embodiments of the invention. Referring to fig. 3, in some embodiments thereof, the generator includes a feature extractor, a first encoder, a second encoder, and a decoder.
Step S120, inputting the real face image and the configuration information thereof into a generator, obtaining a generated face image corresponding to the real face image by the generator, including:
step S121, inputting the real face image into the feature extractor, and inputting the configuration information into the second encoder with the output of the feature extractor as the input of the first encoder.
Step S122, the outputs of the first encoder and the second encoder are used as the input of a mean value layer and a variance layer, a mean value vector is obtained through the mean value layer, and a variance vector is obtained through the variance layer.
Step S123, potential vectors are obtained according to the mean vector and the variance vector.
Step S124, the potential vector is input into a decoder, and the generated face image is obtained through the decoder.
In particular, the generator functions to create an overall representation of the input image based on a variational automatic encoder architecture having an architecture similar to that of an automatic encoder that includes a dimension encoder that compresses the input image and a decoder that decompresses the input image. The deep neural network model of the present invention encodes the potential vectors using random distribution so that the network can learn the domain, allowing it to be better generalized, the generator accepting both color image and configuration information inputs.
The feature extractor consists of three parallel convolution neural networks without weight sharing; the filter sizes of the three convolutional neural networks are 11×11, 7×7 and 5×5 respectively, each convolutional neural network comprises three layers of first sub-networks, and each layer of first sub-network comprises a convolutional layer, a max pooling layer and a ReLU activation function which are sequentially connected.
The first encoder and the second encoder each comprise three layers of second subnetworks; the third layer of second subnetwork is respectively provided with 4096, 2048 and 1024 neurons, and is connected with the mean layer and the variance layer; the mean and variance layers each have 64 neurons.
The decoder comprises three layers of third subnetworks; the three-layer third subnetwork has 1024, 2048 and 4096 neurons, respectively.
Illustratively, the specific workflow of the generator is:
The 128×128 pixel real face image is connected and fed to a first encoder of a 3-layer network structure having 4096, 2048, and 1024 neurons through an output of a feature extractor consisting of 3 convolutional neural networks, a third layer network of the first encoder is connected to a mean layer and a variance layer, configuration information is fed to a second encoder of a 3-layer network structure having 4096, 2048, and 1024 neurons (the structure is the same as the first encoder), and a third layer network of the second encoder is connected to the mean layer and the variance layer. The outputs of the two encoders are connected together to form a mean vector and a variance vector by a mean layer and a variance layer, each vector having 128 elements, the mean vector and the variance vector yielding potential vectors, a specific formula can be expressed as: z=μ+ε·n (0, 1). Where μ represents the mean vector, ε represents the variance vector, and N (0, 1) is the zero mean and unit variance Gaussian distribution. The 128 potential vectors Z are fed to a decoder of the 3-layer network structure, followed by an S-shaped activation function, the expected output of the generator being the frontal face view of the object and the neutral expression of the face (generating a face image).
Further, the classifier comprises a hidden layer, a hyperbolic tangent activation function and an output layer for class prediction which are connected in sequence; the output layer uses Softmax activation to represent the confidence level of each class prediction.
Specifically, the generated face image output by the generator is flattened and fed to a classifier, which is a neural network with one hidden layer of 256 neurons, a hyperbolic tangent activation function, and an output layer for class prediction, which uses Softmax activation to represent the confidence level of each class prediction, with the probability sum of all outputs equal to 1.
In the embodiment, the face recognition task is definitely divided into two subtasks of the generator and the classifier, and the method has the advantages that the extracted facial features reflect the similarity of the target field, and the inside of the deep neural network is mainly driven by data, so that the performance of the deep neural network on invisible data is improved, and the face recognition accuracy of the security face recognition model is further improved.
The security face recognition model needs to be trained before use, and the loss function of the model during training is as follows:
where z represents the potential vector, x represents the data point, μ represents the mean vector, ε represents the variance vector.
Specifically, the loss function is mainly composed of two terms. The first term is the marginal likelihood of the data point, which is the data point that is input by reconstructing from the potential vector by minimizing the term, as a loss of data fidelity. The second term is KL divergence, which maximizes KL divergence, approximating a potential vector.
Because the invention aims to reconstruct the true image data points which are possibly different from the input image, the marginal likelihood and the minimum absolute difference function are used for replacing the data fidelity item, and the face recognition accuracy of the security face recognition model is improved.
The embodiment of the invention also provides a security face recognition device based on deep learning, which is used for recognizing the face through the trained security face recognition model, wherein the security face recognition model comprises a generator and a classifier. The device is used for realizing the above embodiments and preferred embodiments, and is not described in detail. The terms "module," "unit," "sub-unit," and the like as used below may refer to a combination of software and/or hardware that performs a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementations in hardware, or a combination of software and hardware, are also possible and contemplated.
Fig. 4 is a block diagram of a security face recognition device based on deep learning according to an embodiment of the present invention. Referring to fig. 4, the security face recognition apparatus based on deep learning includes:
the data acquisition module is used for acquiring a real face image to be identified and configuration information thereof, wherein the configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image;
The image generation module is used for inputting the real face image and the configuration information thereof into the generator, and obtaining a generated face image corresponding to the real face image through the generator;
The image classification module is used for inputting the generated face image into a classifier, and obtaining the recognition classification result of the real face image through the classifier.
In the technical scheme, the face recognition is carried out on the real face image mainly through the security face recognition model based on deep learning. In addition to taking the real face image as the input of the security face recognition model, the configuration information corresponding to the real face image is also taken as the input of the security face recognition model. The configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image. This allows the network to maximize the pairwise correlation between the front and contour faces and project the data to a common embedding. By the configuration information, the influence of the pose and the expression of the face in the real face image on the face recognition can be relieved, so that the face recognition accuracy is improved, and the problem of low face recognition accuracy in the existing security face recognition method based on deep learning is solved.
The above-described respective modules may be functional modules or program modules, and may be implemented by software or hardware. For modules implemented in hardware, the various modules described above may be located in the same processor; or the above modules may be located in different processors in any combination.
It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to be limiting. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure in accordance with the embodiments provided herein.
It is to be understood that the drawings are merely illustrative of some embodiments of the present application and that it is possible for those skilled in the art to adapt the present application to other similar situations without the need for inventive work. In addition, it should be appreciated that while the development effort might be complex and lengthy, it would nevertheless be a routine undertaking of design, fabrication, or manufacture for those of ordinary skill having the benefit of this disclosure, and thus should not be construed as a departure from the disclosure.
Claims (10)
1. The security face recognition method based on deep learning is characterized by comprising the following steps of:
Acquiring a real face image to be identified and configuration information thereof, wherein the configuration information comprises the distance characteristics between any two parts in the five sense organs of the real face image;
inputting the real face image and configuration information thereof into the generator, and obtaining a generated face image corresponding to the real face image through the generator;
and inputting the generated face image into the classifier, and obtaining a recognition classification result of the real face image through the classifier.
2. The security face recognition method based on deep learning according to claim 1, wherein the obtaining the real face image to be recognized and the configuration information thereof includes:
acquiring a real face image to be identified;
normalizing the illumination variation in the real face image;
and adjusting the normalized pixel intensity value of the real face image by adopting gamma correction.
3. The security face recognition method based on deep learning according to claim 2, wherein the obtaining the real face image to be recognized and the configuration information thereof further comprises:
and positioning the five sense organs in the real face image by adopting a multi-task cascade convolutional neural network, and calculating the distance characteristics between any two parts in the five sense organs to obtain the configuration information of the real face image, wherein the five sense organs comprise eyebrows, eyes, nose, mouth and chin.
4. The depth learning based security face recognition method of claim 1, wherein the generator comprises a feature extractor, a first encoder, a second encoder, and a decoder;
Inputting the real face image and the configuration information thereof into the generator, obtaining a generated face image corresponding to the real face image through the generator, and comprising the following steps:
Inputting the real face image into the feature extractor, taking the output of the feature extractor as the input of the first encoder, and inputting the configuration information into the second encoder;
Taking the output of the first encoder and the output of the second encoder as the input of a mean value layer and a variance layer, obtaining a mean value vector through the mean value layer, and obtaining a variance vector through the variance layer;
Obtaining potential vectors according to the mean vector and the variance vector;
and inputting the potential vector into the decoder, and obtaining the generated face image through the decoder.
5. The deep learning-based security face recognition method of claim 4, wherein the feature extractor consists of three parallel convolutional neural networks without weight sharing;
The filter sizes of the three convolutional neural networks are 11×11, 7×7 and 5×5 respectively, each convolutional neural network comprises three layers of first sub-networks, and each layer of first sub-network comprises a convolutional layer, a maximum pooling layer and a ReLU activation function which are sequentially connected.
6. The depth learning based security face recognition method of claim 4, wherein the first encoder and the second encoder each comprise three layers of second subnetworks;
the third layer of the second sub-network is respectively provided with 4096, 2048 and 1024 neurons, and the third layer of the second sub-network is connected with the mean layer and the variance layer;
the mean layer and the variance layer each have 64 neurons.
7. The deep learning-based security face recognition method of claim 4, wherein the decoder comprises three layers of third subnetworks;
the third subnetwork of three layers has 1024, 2048 and 4096 neurons, respectively.
8. The deep learning-based security face recognition method of claim 1, wherein the classifier comprises a hidden layer, a hyperbolic tangent activation function and an output layer for class prediction which are connected in sequence;
The output layer uses Softmax activation to represent the confidence level of each class prediction.
9. The deep learning-based security face recognition method of claim 4, wherein the loss function of the security face recognition model is:
where z represents the potential vector, x represents the data point, μ represents the mean vector, ε represents the variance vector.
10. Security face recognition device based on degree of depth study, discerns the face through the security face recognition model after the training, security face recognition model includes generator and classifier, its characterized in that, security face recognition device includes:
The data acquisition module is used for acquiring a real face image to be identified and configuration information thereof, wherein the configuration information comprises distance features between any two parts in the five sense organs of the real face image;
The image generation module is used for inputting the real face image and the configuration information thereof into the generator, and obtaining a generated face image corresponding to the real face image through the generator;
and the image classification module is used for inputting the generated face image into the classifier, and obtaining the recognition classification result of the real face image through the classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410248041.6A CN117994835A (en) | 2024-03-05 | 2024-03-05 | Security face recognition method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410248041.6A CN117994835A (en) | 2024-03-05 | 2024-03-05 | Security face recognition method and device based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117994835A true CN117994835A (en) | 2024-05-07 |
Family
ID=90897511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410248041.6A Pending CN117994835A (en) | 2024-03-05 | 2024-03-05 | Security face recognition method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117994835A (en) |
-
2024
- 2024-03-05 CN CN202410248041.6A patent/CN117994835A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112766160B (en) | Face replacement method based on multi-stage attribute encoder and attention mechanism | |
| Yuan et al. | Fingerprint liveness detection using an improved CNN with image scale equalization | |
| CN108537743A (en) | A kind of face-image Enhancement Method based on generation confrontation network | |
| Tarrés et al. | A novel method for face recognition under partial occlusion or facial expression variations | |
| Lin et al. | Near-realtime face mask wearing recognition based on deep learning | |
| CN113705290A (en) | Image processing method, image processing device, computer equipment and storage medium | |
| CN116343330A (en) | Abnormal behavior identification method for infrared-visible light image fusion | |
| CN113537027B (en) | Face depth counterfeiting detection method and system based on face division | |
| CN112418041A (en) | Multi-pose face recognition method based on face orthogonalization | |
| CN110490057B (en) | Self-adaptive identification method and system based on human face big data artificial intelligence clustering | |
| Liu et al. | Physics-guided spoof trace disentanglement for generic face anti-spoofing | |
| CN116012255A (en) | Low-light image enhancement method for generating countermeasure network based on cyclic consistency | |
| CN110543813B (en) | Face image and gaze counting method and system based on scene | |
| Garg et al. | Facial expression recognition & classification using hybridization of ICA, GA, and neural network for human-computer interaction | |
| CN112395971A (en) | StarGAN-based face test image generation method under different quantitative illumination and angle conditions, application and storage medium | |
| CN114862716B (en) | Image enhancement method, device, equipment and storage medium for face image | |
| Zhao et al. | Cross-view gait recognition based on dual-stream network | |
| Shukla et al. | Deep Learning Model to Identify Hide Images using CNN Algorithm | |
| CN110135362A (en) | A kind of fast face recognition method based under infrared camera | |
| CN117994835A (en) | Security face recognition method and device based on deep learning | |
| Nasiri et al. | Video Surveillance Framework Based on Real-Time Face Mask Detection and Recognition | |
| CN112818808B (en) | High-precision gait recognition method combining two vector embedding spaces | |
| Shu et al. | The research and implementation of human posture recognition algorithm via OpenPose | |
| Chen | Design and simulation of AI remote terminal user identity recognition system based on reinforcement learning | |
| CN113553895A (en) | Multi-pose face recognition method based on face orthogonalization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |