WO2020233000A1

WO2020233000A1 - Facial recognition method and apparatus, and computer-readable storage medium

Info

Publication number: WO2020233000A1
Application number: PCT/CN2019/117342
Authority: WO
Inventors: 刘洋; 陈海平
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-05-20
Filing date: 2019-11-12
Publication date: 2020-11-26
Also published as: CN110309709A

Abstract

The present application relates to an artificial intelligence technology. Disclosed is a facial recognition method, comprising: collecting an original data set and a facial contrast set, pre-processing same and then inputting same into a model training layer; the model training layer performing calculation according to a histogram of oriented gradients method to obtain a gradient feature set; inputting the gradient feature set and the facial contrast set into a lifting algorithm for training, and exiting the training when the training accuracy of the lifting algorithm is greater than a pre-set threshold value; receiving a captured image, and the model training layer determining whether the captured image contains a face; and when the captured image contains a face, searching for a face with the highest similarity to the facial contrast set to complete facial recognition. Further provided are a facial recognition apparatus and a computer-readable storage medium. The present application can realize an accurate facial recognition function.

Description

Face recognition method, device and computer readable storage medium

Based on the Paris Convention, this application declares that it enjoys the priority of the Chinese patent application filed on May 20, 2019, with the application number CN201910417997.3 and the title "face recognition method, device and computer readable storage medium". The Chinese patent application The overall content of is incorporated in this application by reference.

Technical field

This application relates to the field of artificial intelligence technology, and in particular to a face recognition method, device and computer-readable storage medium that can be used for smart security.

Background technique

With the advancement of science and technology and the continuous improvement of people’s living standards, people’s requirements for the living and working environment are becoming more and more stringent, especially for safety and intelligence. The video surveillance system is the representative system of security. With the continuous development of Internet of Things technology, the importance of video surveillance systems in the field of national security and urbanization management has become more and more prominent, and the requirements for its functions and performance have also been continuously improved. However, due to the massive increase of the migrant population in today's society, the lack of information management methods and many other factors, traffic accidents and social security problems are showing increasingly serious situations. According to the public security department, criminal activities such as theft and robbery by breaking doors or breaking windows into experimental and scientific research buildings, office buildings, and residential communities are still very serious. Therefore, the improvement and development of security monitoring systems are urgent tasks. At present, the security monitoring based on face recognition at home and abroad mainly adopts the face alarm system of infrared sensor, but the infrared alarm system is susceptible to interference from various heat sources, light sources, radio frequency radiation, and hot air flow, and it is difficult to achieve efficient face recognition effects.

Summary of the invention

This application provides a face recognition method, device, and computer-readable storage medium, the main purpose of which is to provide a technical solution that can efficiently recognize a face from video or picture data.

In order to achieve the above objective, a face recognition method provided by this application includes:

Step A: The data collection layer collects a face image set, a non-face image set, and a face comparison set, saves the face image set and the non-face image set as an original data set, and sends the original data Input the set to the data processing layer, and input the face comparison set into the database;

Step B: The data processing layer performs grayscale and denoising processing on the original data set to obtain a preprocessed data set, where the preprocessed data set includes a face preprocessing data set and a non-face preprocessing Data set, input the face preprocessing data set to the data cutting layer, and input the non-face preprocessing data set to the model training layer;

Step C: The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set, and then obtains the face training set and inputs it to the model training layer;

Step D: The model training layer receives a training set consisting of the face training set and the non-face preprocessing data set, and extracts the face control set from the database, according to the direction gradient straight method The training set is calculated to obtain a gradient feature set, and the gradient feature set and the face control set are input to the lifting algorithm for training, until the training accuracy of the lifting algorithm is greater than a preset threshold, the model The training layer exits training;

Step E: The data acquisition layer receives the captured image, performs grayscale and noise reduction processing on the input image, and then inputs it to the model training layer. The model training layer determines whether the captured image contains a person Face, when the captured image does not contain a human face, output the result that the human face is not recognized;

Step F: When the captured image contains a human face, the model training layer sequentially determines the similarity between the captured image and the face control set of the database based on the Euclidean distance method, and outputs the highest similarity Face comparison set pictures to complete face recognition.

In addition, in order to achieve the above object, the present application also provides a face recognition device, which includes a memory and a processor. The memory stores a face recognition program that can run on the processor. The following steps are implemented when the recognition program is executed by the processor:

Step D: The model training layer receives a training set consisting of the face training set and the non-face preprocessing data set, and extracts the face control set from the database, according to the direction gradient straight method Calculate the training set to obtain a gradient feature set, and input the gradient feature set and the face control set to the lifting algorithm for training, until the training accuracy of the lifting algorithm is greater than a preset threshold, the model training layer Exit training;

Step E: The data acquisition layer receives the captured image, performs grayscale and noise reduction processing on the captured image, and then inputs it to the model training layer. The model training layer determines whether the captured image is Contains a human face, and when the captured image does not contain a human face, output the result that the human face is not recognized;

In addition, in order to achieve the above-mentioned object, the present application also provides a computer-readable storage medium that stores a face recognition program on the computer-readable storage medium, and the face recognition program can be executed by one or more processors to Implement the steps of the face recognition method as described above.

The adaptive image denoising filter can reduce the impact of noise on the image, and the improvement algorithm makes good use of weak classifiers for cascading, and the final combination form of strong classifiers has high classification accuracy. Therefore, the face recognition proposed in this application The method, device and computer-readable storage medium can realize accurate face recognition function.

Description of the drawings

FIG. 1 is a schematic flowchart of a face recognition method provided by an embodiment of this application;

2 is a schematic diagram of the internal structure of a face recognition device provided by an embodiment of the application;

FIG. 3 is a schematic diagram of modules of a face recognition program in a face recognition device provided by an embodiment of the application.

The realization, functional characteristics, and advantages of the purpose of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described here are only used to explain the application, and are not used to limit the application.

This application provides a face recognition method. Referring to FIG. 1, it is a schematic flowchart of a face recognition method provided by an embodiment of this application. The method can be executed by a device, and the device can be implemented by software and/or hardware.

In this embodiment, the face recognition method includes:

S1. The data collection layer collects a face image set, a non-face image set, and a face comparison set, saves the face image set and the non-face image set as an original data set, and sends the original data set Input to the data processing layer, and input the face comparison set into the database.

The preferred embodiment of the present application deploys several video surveillance areas in a preset scene, such as an experimental scientific research building, office building, residential area, etc., and selects images including human faces from the images captured in the several video surveillance areas. Form a face image set; based on different faces in the face image set, collect ID photo pictures corresponding to the different faces, and obtain ID photos from the relevant monitoring department, such as from the public security department The ID photos of the criminals at large, the ID photos of the untrustworthy Lai Lai, etc., constitute the face comparison set.

Further, the preferred embodiment of the present application selects images that do not include human faces from the captured image sets in the several video surveillance areas, and obtains non-human target data sets from a preset data set, such as the COCO data set , Compose a set of non-face images. The COCO data set is a large-scale image data set specially designed for object detection, segmentation, human key point detection, semantic segmentation and caption generation.

S2. The data processing layer performs grayscale and denoising processing on the original data set to obtain a preprocessed data set, where the preprocessed data set includes a face preprocessed data set and a non-face preprocessed data Set, input the face preprocessing data set to the data cutting layer, and input the non-face preprocessing data set to the model training layer.

In a preferred embodiment of the present application, the gray scale is to convert the data in the original data set from an RGB format to a black and white gray format by using a proportional method. The ratio method is as follows: Obtain the R, G, and B pixel values of each pixel in the original data set, and convert the pixel to a black-and-white gray format according to the following function:

0.30*R+0.59*G+0.11*B

In a preferred embodiment of the present application, the noise reduction processing adopts the following adaptive image noise reduction filtering method:

g(x,y)=η(x,y)+f(x,y)

Wherein, (x, y) represents the coordinates of the image pixels in the original data set, and f(x, y) is the output data after the original data set is denoised based on the adaptive image noise reduction filtering method , Η(x,y) is noise, g(x,y) is the original data set,

Is the total noise variance of the original data set,

Is the average gray value of the pixel (x, y),

Is the pixel gray variance of the (x, y), and L represents the current pixel coordinates.

S3. The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set to obtain a face training set and inputs it to the model training layer.

In the preferred embodiment of the present application, the edge detection and the segmentation process are based on the edge detection to find a pixel set with a large change in the pixel gray level in the face preprocessing data set, and based on the segmentation process. The pixel set is reconnected to segment the human face and the human face background. Wherein, the larger step change means that the gray-scale derivative has a maximum value or a minimum value.

The preferred embodiment of the present application adopts the Canny edge detection method. The Canny edge detection method smooths and filters the face preprocessing data set based on a Gaussian filter, and calculates the smoothed and filtered data set based on the first-order partial derivative finite difference to obtain non-local maximum points and Minimal value point, complete edge detection.

S4. The model training layer receives a training set consisting of the face training set and the non-face preprocessing data set, and extracts the face control set from the database, and calculates according to the direction gradient straight method The training set obtains a gradient feature set, and the gradient feature set and the face control set are input to a lifting algorithm for training, until the training accuracy of the lifting algorithm is greater than a preset threshold, the model training layer exits training.

The preferred embodiment of the present application calculates the gradient amplitude and gradient direction value of each pixel (x, y) of the data in the training set, and uses the gradient amplitude as the first component and the gradient direction value as the first component. Two components form a gradient matrix; divide the data in the gradient matrix into multiple small blocks, and add the gradient amplitude and the gradient direction value of each small block to obtain the added value, and the added value is connected in series to form the The gradient feature set.

In a preferred embodiment of the present application, the boosting algorithm includes the AdaBoost algorithm, and the AdaBoost algorithm includes several weak classifiers and strong classifiers;

Wherein, the weak classifier h(x, t, p, θ) is:

Wherein, t is the classification function including the gradient feature set, x is the detection sub-window, p is the weighted inequality direction coefficient, and θ is the weak classifier threshold. The preferred embodiment of the present application is trained on the basis of the gradient feature set. The weak classifier h(x, t, p, θ) until the optimal threshold θ is determined to obtain the strong classifier C(x):

Where α _k is the coefficient of the strong classifier C(x), T is the total number of the weak classifiers, β _k =ε _k /(1-ε _k ), and the ε _k is:

Wherein, w _{i is} the weight of the gradient feature set, and _yi is the face control set.

In a preferred embodiment of the present application, when the strong classifier C(x) judges that the accuracy of the face recognition of the training set is greater than the preset threshold, the boosting algorithm exits training, and the preset threshold is generally Set to 0.97.

S5. Receive a captured image, perform grayscale and noise reduction processing on the captured image, and then input it to the model training layer to determine whether the captured image contains a human face.

Preferably, the captured image uses an image captured by equipment such as an outdoor camera and a mobile phone.

S6. When the captured image does not contain a human face, output the result that the human face is not recognized.

S7. When the captured image contains a face, the model training layer sequentially determines the similarity between the captured image and the face comparison set of the database based on the Euclidean distance method, and outputs the face comparison with the highest similarity Collect pictures to complete face recognition.

In a preferred embodiment of the present application, the Euclidean distance method is:

Wherein, a is the captured image, _yi is the face comparison set, and n is the total amount of data in the face comparison set.

The invention also provides a face recognition device. Referring to FIG. 2, it is a schematic diagram of the internal structure of a face recognition device provided by an embodiment of this application. (Corresponding modification)

In this embodiment, the face recognition apparatus 1 may be a PC (Personal Computer, personal computer), or a terminal device such as a smart phone, a tablet computer, or a portable computer, or a server. The face recognition device 1 at least includes a memory 11, a processor 12, a communication bus 13, and a network interface 14.

Wherein, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. The memory 11 may be an internal storage unit of the face recognition device 1 in some embodiments, such as a hard disk of the face recognition device 1. In other embodiments, the memory 11 may also be an external storage device of the face recognition device 1, for example, a plug-in hard disk equipped on the face recognition device 1, a smart memory card (Smart Media Card, SMC), and a secure digital (Secure Digital) Digital, SD) card, flash card (Flash Card), etc. Further, the memory 11 may also include both an internal storage unit of the face recognition apparatus 1 and an external storage device. The memory 11 can be used not only to store application software and various data installed in the face recognition device 1, such as the code of the face recognition program 01, etc., but also to temporarily store data that has been output or will be output.

In some embodiments, the processor 12 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip, and is used to run the program code or processing stored in the memory 11 Data, such as execution of face recognition program 01, etc.

The communication bus 13 is used to realize the connection and communication between these components.

The network interface 14 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface), and is usually used to establish a communication connection between the device 1 and other electronic devices.

Optionally, the device 1 may also include a user interface. The user interface may include a display (Display) and an input unit such as a keyboard (Keyboard). The optional user interface may also include a standard wired interface and a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light emitting diode) touch device, etc. Among them, the display can also be called a display screen or a display unit as appropriate, for displaying information processed in the face recognition device 1 and for displaying a visualized user interface.

FIG. 2 only shows the face recognition device 1 with components 11-14 and the face recognition program 01. Those skilled in the art will understand that the structure shown in FIG. 1 does not constitute a limitation on the face recognition device 1 It may include fewer or more components than shown, or a combination of some components, or a different component arrangement.

In the embodiment of the device 1 shown in FIG. 2, the memory 11 stores the face recognition program 01; when the processor 12 executes the face recognition program 01 stored in the memory 11, the following steps are implemented:

Step 1. Collect a face image set, a non-face image set, and a face comparison set. The face image set and the non-face image set are collectively referred to as the original data set, and the original data set is input to the data processing Layer, input the face comparison set into the database.

The preferred embodiment of the present application deploys several video surveillance areas, and transfers the image sets captured in the several video surveillance areas to the database; selects images that include human faces in the image set stored in the database to form human faces Image set; based on the different faces in the face image set, collect ID photos corresponding to the different faces, and obtain ID photos from the relevant monitoring department, such as obtaining criminals at large from the public security department The ID photos and the ID photos of the untrustworthy old Lai, etc. form the face comparison set.

The preferred embodiment of this application selects images that do not include human faces in the image set stored in the database, and selects non-human target data sets from a preset data set, such as the COCO data set, to form a non-human face image set. The COCO data set is a large-scale image data set designed for object detection, segmentation, human key point detection, semantic segmentation and caption generation.

Step 2: The data processing layer performs grayscale and denoising processing on the original data set to obtain a preprocessed data set, and the preprocessed data set includes a face preprocessed data set and a non-face preprocessed data set, The face preprocessing data set is input to the data cutting layer, and the non-face preprocessing data set is input to the model training layer.

0.30*R+0.59*G+0.11*B

The noise reduction processing in the preferred embodiment of the application adopts an adaptive image noise reduction filtering method:

g(x,y)=η(x,y)+f(x,y)

Is the total noise variance of the original data set,

Is the average gray value of the pixel (x, y),

Step 3. The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set, and then obtains the face training set and inputs it to the model training layer. The face training set The non-face preprocessing data set is collectively referred to as a training set.

In the preferred embodiment of the present application, the edge detection and the segmentation process are based on the edge detection to find a pixel set with a large change in the pixel gray level in the face preprocessing data set, and based on the segmentation process. The pixel set is reconnected to segment the human face and the human face background. Further, the step change is large, that is, the gray-scale derivative is a maximum value or a minimum value.

The preferred embodiment of the present application adopts the Canny edge detection method. The Canny edge detection method performs smoothing filtering processing on the face preprocessing data set based on a Gaussian filter, and calculating after the smoothing filtering processing based on the first-order partial derivative finite difference Obtain the non-local maximum and minimum points for the data set, and complete the edge detection.

Step 4. The model training layer receives the training set and extracts the face control set from the database, calculates the training set according to the directional gradient method to obtain a gradient feature set, and compares the gradient feature set with all The face comparison set is input to the lifting algorithm for training, and training is exited when the training accuracy of the lifting algorithm is greater than a preset threshold.

In a preferred embodiment of this application, the gradient amplitude and gradient direction value of each pixel (x, y) of the data in the training set are calculated, and the gradient amplitude is used as the first component, and the gradient direction value is used as The second component forms a gradient matrix; divide the data in the gradient matrix into a plurality of small blocks, and add the gradient amplitude and the gradient direction value of each small block to obtain an added value, and connect the added value in series to form a gradient The feature set is input to the lifting algorithm.

The boosting algorithm of the preferred embodiment of the present application includes the AdaBoost algorithm, and the AdaBoost algorithm includes several weak classifiers and strong classifiers;

The weak classifier h(x, t, p, θ) is:

Where t is the classification function including the gradient feature set, x is the detection sub-window, p is the weighted inequality direction coefficient, and θ is the threshold of the weak classifier, and the weak classifier is trained according to the gradient feature set. Weak classifier h(x,t,p,θ) until the optimal threshold θ is determined;

The strong classifier C(x) is:

Step 5: Receive a captured image, perform grayscale and noise reduction processing on the captured image, and input it to the model training layer to determine whether the captured image contains a human face.

Step 6. When the captured image does not contain a human face, output the result that the human face is not recognized.

Step 7. When the captured image contains a face, the model training layer sequentially judges the similarity between the captured image and the face comparison set of the database based on the Euclidean distance method, and outputs the face with the highest similarity Comparing the pictures in the collection to complete face recognition.

Optionally, in other embodiments, the face recognition program can also be divided into one or more modules, and the one or more modules are stored in the memory 11 and run by one or more processors (this embodiment is The processor 12) is executed to complete the application. The module referred to in the application refers to a series of computer program instruction segments capable of completing specific functions, and is used to describe the execution process of the face recognition program in the face recognition device.

For example, referring to FIG. 3, it is a schematic diagram of the program modules of the face recognition program in an embodiment of the applicant’s face recognition device. In this embodiment, the face recognition program can be divided into a data receiving module 10 and a data The processing module 20, the model training module 30, and the face recognition output module 40 are exemplary:

The data receiving module 10 is used to collect a face image set, a non-face image set, and a face comparison set. The face image set and the non-face image set are collectively referred to as an original data set, and the original The data set is input into the data processing layer, and the face comparison set is input into the database.

The data processing module 20 is configured to: the data processing layer performs grayscale and noise reduction processing on the original data set to obtain a preprocessed data set, and the preprocessed data set includes a face preprocessed data set and a non-human face The face preprocessing data set is input to the data cutting layer, and the non-face preprocessing data set is input to the model training layer. The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set, and then obtains the face training set and inputs it to the model training layer. The face training set and the Non-face preprocessing data sets are collectively referred to as training sets.

The model training module 30 is configured to: the model training layer receives the training set and extracts the face control set from the database, calculates the training set according to the directional gradient method to obtain a gradient feature set, The gradient feature set and the face control set are input to a lifting algorithm for training, and training is exited when the training accuracy of the lifting algorithm is greater than a preset threshold.

The face recognition output module 40 is configured to: receive the captured image, perform grayscale and noise reduction processing on the image captured by the user, and then input it to the model training layer, and the model training layer determines the captured image Whether the captured image contains a human face, when the captured image does not contain a human face, output the result that the human face is not recognized. When the captured image contains a face, the model training layer sequentially determines the similarity between the captured image and the face comparison set of the database based on the Euclidean distance method, and outputs the face comparison with the highest similarity Collect pictures to complete face recognition.

The functions or operation steps implemented by the program modules such as the data receiving module 10, the data processing module 20, the model training module 30, and the face recognition output module 40 when executed are substantially the same as those in the foregoing embodiment, and will not be repeated here.

In addition, an embodiment of the present application also proposes a computer-readable storage medium that stores a face recognition program on the computer-readable storage medium, and the face recognition program can be executed by one or more processors to implement the following operations :

Collect a face image set, a non-face image set, and a face control set. The face image set and the non-face image set are collectively referred to as the original data set. The original data set is input to the data processing layer, and The face comparison set is input into the database.

The data processing layer performs grayscale and noise reduction processing on the original data set to obtain a preprocessed data set. The preprocessed data set includes a face preprocessed data set and a non-face preprocessed data set, and the The face preprocessing data set is input to the data cutting layer, and the non-face preprocessing data set is input to the model training layer. The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set, and then obtains the face training set and inputs it to the model training layer. The face training set and the Non-face preprocessing data sets are collectively referred to as training sets.

The model training layer receives the training set and extracts the face control set from the database, calculates the training set according to the directional gradient method to obtain a gradient feature set, and compares the gradient feature set with the face The control set is input to the lifting algorithm for training, and the training is exited when the training accuracy of the lifting algorithm is greater than the preset threshold.

Receive the captured image, perform grayscale and noise reduction processing on the captured image, and then input it to the model training layer. The model training layer determines whether the captured image contains a human face. When the captured image is The obtained image does not contain human faces, and the result of unrecognized human faces is output. When the captured image contains a face, the model training layer sequentially determines the similarity between the captured image and the face comparison set of the database based on the Euclidean distance method, and outputs the face comparison with the highest similarity Collect pictures to complete face recognition.

It should be noted that the serial numbers of the above embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. And the terms "include", "include" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes The other elements listed may also include elements inherent to the process, device, article, or method. If there are no more restrictions, the element defined by the sentence "including a..." does not exclude the existence of other identical elements in the process, device, article or method that includes the element.

Through the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better.的实施方式。 Based on this understanding, the technical solution of this application essentially or the part that contributes to the existing technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM/RAM) as described above. , Magnetic disk, optical disk), including several instructions to make a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) execute the method described in each embodiment of the present application.

The above are only preferred embodiments of this application, and do not limit the scope of this application. Any equivalent structure or equivalent process transformation made using the content of the description and drawings of this application, or directly or indirectly used in other related technical fields , The same reason is included in the scope of patent protection of this application.

Claims

A face recognition method, characterized in that the method includes:

Step A: The data collection layer collects a face image set, a non-face image set, and a face comparison set, saves the face image set and the non-face image set as an original data set, and sends the original data Input the set to the data processing layer, and input the face comparison set into the database;

Step B: The data processing layer performs grayscale and denoising processing on the original data set to obtain a preprocessed data set, where the preprocessed data set includes a face preprocessing data set and a non-face preprocessing Data set, input the face preprocessing data set to the data cutting layer, and input the non-face preprocessing data set to the model training layer;

Step C: The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set to obtain a face training set and inputs it to the model training layer;

Step D: The model training layer receives a training set consisting of the face training set and the non-face preprocessing data set, and extracts the face control set from the database, according to the direction gradient straight method Calculate the training set to obtain a gradient feature set, and input the gradient feature set and the face control set to the lifting algorithm for training, until the training accuracy of the lifting algorithm is greater than a preset threshold, the model training layer Exit training;

Step E: The data acquisition layer receives the captured image, performs grayscale and noise reduction processing on the captured image, and then inputs it to the model training layer. The model training layer determines whether the captured image is Contains a human face, and when the captured image does not contain a human face, output the result that the human face is not recognized;

Step F: When the captured image contains a human face, the model training layer sequentially determines the similarity between the captured image and the face control set of the database based on the Euclidean distance method, and outputs the highest similarity Face comparison set pictures to complete face recognition.
The face recognition method according to claim 1, wherein collecting a face image set, a non-face image set and a face comparison set comprises:

Select images including faces from a collection of images captured in several video surveillance areas deployed in preset scenes to form a face image set;

Select images that do not include human faces from the captured image set, and select non-human target data sets from the preset database to form a non-human face image set;

Based on different faces in the face image set, ID photo images corresponding to the different faces are collected to form a face comparison set.
The face recognition method of claim 1 or 2, wherein the noise reduction processing adopts the following adaptive image noise reduction filtering method:

g(x,y)=η(x,y)+f(x,y)

Wherein, (x, y) represents the coordinates of the image pixels in the original data set, and f(x, y) is the output data after the original data set is denoised based on the adaptive image noise reduction filtering method , Η(x,y) is noise, g(x,y) is the original data set,
Is the total noise variance of the original data set,
Is the average gray value of the pixel (x, y),
Is the pixel gray variance of the (x, y), and L represents the current pixel coordinates.
The face recognition method according to claim 3, wherein the edge detection adopts the Canny edge detection method.
The face recognition method according to claim 3, wherein the calculating the training set according to the directional gradient straight method to obtain the gradient feature set comprises:

Calculate the gradient amplitude and gradient direction value of each pixel (x, y) of the data in the training set, and use the gradient amplitude as the first component and the gradient direction value as the second component to form a gradient matrix;

The data in the gradient matrix is divided into a plurality of small blocks, and the gradient amplitude and the gradient direction value of each small block are added to obtain an added value, and the added value is connected in series to form the gradient feature set.
The face recognition method according to claim 5, wherein the boosting algorithm comprises an AdaBoost algorithm, and the AdaBoost algorithm comprises several weak classifiers and strong classifiers;

Wherein, the weak classifier h(x, t, p, θ) is:

Where t is the classification function including the gradient feature set, x is the detection sub-window, p is the weighted inequality direction coefficient, and θ is the threshold of the weak classifier, and the weak classifier is trained according to the gradient feature set. The weak classifier h(x, t, p, θ) until the optimal threshold θ is determined to obtain the strong classifier C(x):

Where α k is the coefficient of the strong classifier C(x), T is the total number of the weak classifiers, β k =ε k /(1-ε k ), and the ε k is:

Wherein, w i is the weight of the gradient feature set, and yi is the face control set.
8. The face recognition method of claim 6, wherein the Euclidean distance method is:

Among them, a is the captured image, yi is the face comparison set, and n is the total amount of data in the face comparison set.
A face recognition device, characterized in that the device includes a memory and a processor, the memory stores a face recognition program that can be run on the processor, and the face recognition program is processed by the processor. The following steps are implemented when the device is executed:

Step A: The data collection layer collects a face image set, a non-face image set, and a face comparison set, saves the face image set and the non-face image set as an original data set, and sends the original data Input the set to the data processing layer, and input the face comparison set into the database;

Step B: The data processing layer performs grayscale and denoising processing on the original data set to obtain a preprocessed data set, where the preprocessed data set includes a face preprocessing data set and a non-face preprocessing Data set, input the face preprocessing data set to the data cutting layer, and input the non-face preprocessing data set to the model training layer;

Step C: The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set, and then obtains the face training set and inputs it to the model training layer;

Step D: The model training layer receives a training set consisting of the face training set and the non-face preprocessing data set, and extracts the face control set from the database, according to the direction gradient straight method Calculate the training set to obtain a gradient feature set, and input the gradient feature set and the face control set to the lifting algorithm for training, until the training accuracy of the lifting algorithm is greater than a preset threshold, the model training layer Exit training;

Step E: The data acquisition layer receives the captured image, performs grayscale and noise reduction processing on the captured image, and then inputs it to the model training layer. The model training layer determines whether the captured image is Contains a human face, and when the captured image does not contain a human face, output the result that the human face is not recognized;

Step F: When the captured image contains a human face, the model training layer sequentially determines the similarity between the captured image and the face control set of the database based on the Euclidean distance method, and outputs the highest similarity Face comparison set pictures to complete face recognition.
8. The face recognition device according to claim 8, wherein collecting a face image set, a non-face image set and a face comparison set comprises:

Select images including faces from a collection of images captured in several video surveillance areas deployed in preset scenes to form a face image set;

Select images that do not include human faces from the captured image set, and select non-human target data sets from the preset database to form a non-human face image set;

Based on different faces in the face image set, ID photo images corresponding to the different faces are collected to form a face comparison set.
The face recognition device according to claim 9, wherein the noise reduction processing adopts the following adaptive image noise reduction filtering method:

g(x,y)=η(x,y)+f(x,y)

Wherein, (x, y) represents the coordinates of the image pixels in the original data set, and f(x, y) is the output data after the original data set is denoised based on the adaptive image noise reduction filtering method , Η(x,y) is noise, g(x,y) is the original data set,
Is the total noise variance of the original data set,
Is the average gray value of the pixel (x, y),
Is the pixel gray variance of the (x, y), and L represents the current pixel coordinates.
The face recognition device of claim 10, wherein the edge detection adopts the Canny edge detection method.
The face recognition device according to claim 10, wherein the calculating the training set according to the directional gradient straight method to obtain the gradient feature set comprises:

Calculate the gradient amplitude and gradient direction value of each pixel (x, y) of the data in the training set, and use the gradient amplitude as the first component and the gradient direction value as the second component to form a gradient matrix;

The data in the gradient matrix is divided into a plurality of small blocks, and the gradient amplitude and the gradient direction value of each small block are added to obtain an added value, and the added value is connected in series to form the gradient feature set.
The face recognition device according to claim 10, wherein the boosting algorithm comprises an AdaBoost algorithm, and the AdaBoost algorithm comprises several weak classifiers and strong classifiers;

Wherein, the weak classifier h(x, t, p, θ) is:

Where t is the classification function including the gradient feature set, x is the detection sub-window, p is the weighted inequality direction coefficient, and θ is the threshold of the weak classifier, and the weak classifier is trained according to the gradient feature set. The weak classifier h(x, t, p, θ) until the optimal threshold θ is determined to obtain the strong classifier C(x):

Where α k is the coefficient of the strong classifier C(x), T is the total number of the weak classifiers, β k =ε k /(1-ε k ), and the ε k is:

Wherein, w i is the weight of the gradient feature set, and yi is the face control set.
The face recognition device according to claim 13, wherein the Euclidean distance method is:

Among them, a is the captured image, yi is the face comparison set, and n is the total amount of data in the face comparison set.
A computer-readable storage medium, characterized in that a face recognition program is stored on the computer-readable storage medium, and the face recognition program can be executed by one or more processors to implement the following steps:

Step A: The data collection layer collects a face image set, a non-face image set, and a face comparison set, saves the face image set and the non-face image set as an original data set, and sends the original data Input the set to the data processing layer, and input the face comparison set into the database;

Step B: The data processing layer performs grayscale and denoising processing on the original data set to obtain a preprocessed data set, where the preprocessed data set includes a face preprocessing data set and a non-face preprocessing Data set, input the face preprocessing data set to the data cutting layer, and input the non-face preprocessing data set to the model training layer;

Step C: The data cutting layer receives the face preprocessing data set, performs edge detection and segmentation processing on the face preprocessing data set, and then obtains the face training set and inputs it to the model training layer;

Step D: The model training layer receives a training set consisting of the face training set and the non-face preprocessing data set, and extracts the face control set from the database, according to the direction gradient straight method Calculate the training set to obtain a gradient feature set, and input the gradient feature set and the face control set to the lifting algorithm for training, until the training accuracy of the lifting algorithm is greater than a preset threshold, the model training layer Exit training;

Step E: The data acquisition layer receives the captured image, performs grayscale and noise reduction processing on the captured image, and then inputs it to the model training layer. The model training layer determines whether the captured image is Contains a human face, and when the captured image does not contain a human face, output the result that the human face is not recognized;

Step F: When the captured image contains a human face, the model training layer sequentially determines the similarity between the captured image and the face control set of the database based on the Euclidean distance method, and outputs the highest similarity Face comparison set pictures to complete face recognition.
15. The computer-readable storage medium of claim 15, wherein the collection of a face image set, a non-face image set, and a face comparison set comprises:

Select images including faces from a collection of images captured in several video surveillance areas deployed in preset scenes to form a face image set;

Select images that do not include human faces from the captured image set, and select non-human target data sets from the preset database to form a non-human face image set;

Based on different faces in the face image set, ID photo images corresponding to the different faces are collected to form a face comparison set.
The computer-readable storage medium according to claim 15 or 16, wherein the noise reduction processing adopts the following adaptive image noise reduction filtering method:

g(x,y)=η(x,y)+f(x,y)

Wherein, (x, y) represents the coordinates of the image pixels in the original data set, and f(x, y) is the output data after the original data set is denoised based on the adaptive image noise reduction filtering method , Η(x,y) is noise, g(x,y) is the original data set,
Is the total noise variance of the original data set,
Is the average gray value of the pixel (x, y),
Is the pixel gray variance of the (x, y), and L represents the current pixel coordinates.
17. The computer-readable storage medium of claim 17, wherein the edge detection adopts a Canny edge detection method.
17. The computer-readable storage medium according to claim 17, wherein said calculating said training set to obtain a gradient feature set according to a directional gradient straight method comprises:

Calculate the gradient amplitude and gradient direction value of each pixel (x, y) of the data in the training set, and use the gradient amplitude as the first component and the gradient direction value as the second component to form a gradient matrix;

The data in the gradient matrix is divided into a plurality of small blocks, and the gradient amplitude and the gradient direction value of each small block are added to obtain an added value, and the added value is connected in series to form the gradient feature set.
18. The computer-readable storage medium of claim 19, wherein the boosting algorithm comprises an AdaBoost algorithm, and the AdaBoost algorithm comprises several weak classifiers and strong classifiers;

Wherein, the weak classifier h(x, t, p, θ) is:

Where t is the classification function including the gradient feature set, x is the detection sub-window, p is the weighted inequality direction coefficient, and θ is the threshold of the weak classifier, and the weak classifier is trained according to the gradient feature set. The weak classifier h(x, t, p, θ) until the optimal threshold θ is determined to obtain the strong classifier C(x):

Where α k is the coefficient of the strong classifier C(x), T is the total number of the weak classifiers, β k =ε k /(1-ε k ), and the ε k is:

Wherein, w i is the weight of the gradient feature set, and yi is the face control set.