CN113657187A - Face recognition method, face recognition equipment and computer-readable storage medium - Google Patents
Face recognition method, face recognition equipment and computer-readable storage medium Download PDFInfo
- Publication number
- CN113657187A CN113657187A CN202110845152.1A CN202110845152A CN113657187A CN 113657187 A CN113657187 A CN 113657187A CN 202110845152 A CN202110845152 A CN 202110845152A CN 113657187 A CN113657187 A CN 113657187A
- Authority
- CN
- China
- Prior art keywords
- face
- facial
- target
- recognition
- image
- 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 72
- 230000036544 posture Effects 0.000 claims abstract description 69
- 230000001815 facial effect Effects 0.000 claims description 278
- 238000000605 extraction Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 7
- 238000002372 labelling Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a face recognition method, a face recognition device and a computer-readable storage medium, wherein the method comprises the following steps: by the method, the face recognition is performed by the recognition models with different face postures and the reference face features, so that the limitation on the face posture of the user during face recognition is reduced, and the accuracy of face recognition in different postures is improved.
Description
Technical Field
The present application relates to the field of image processing technologies, and in particular, to a face recognition method, a face recognition device, and a computer-readable storage medium.
Background
The face recognition technology is to distinguish individuals by using the biological features of the person, and usually compare the collected face information with the face information in the base database to obtain a face recognition result.
In a long-term research and development process, the applicant of the present application finds that in the existing face recognition technology, since the collected face gestures are various, such as a front face, a side face, a head drop, etc., and different face gestures are important factors affecting the recognition accuracy, the image of the front face gesture generally contains abundant face information, so that the accuracy of recognition using the image of the front face gesture is high, and the faces of non-front face gestures such as a head drop, a head raising, a side face, etc. are relatively few, so the accuracy of recognition using the non-front face gesture is low.
Therefore, how to improve the accuracy of face recognition in different postures is a key issue in the current face recognition field.
Disclosure of Invention
The technical problem mainly solved by the application is to provide a face recognition method, a face recognition device and a computer-readable storage medium, which can accurately recognize face images in different postures.
In order to solve the technical problem, the application adopts a technical scheme that: there is provided a face recognition method, the method including: carrying out gesture recognition on the target face image by using a gesture recognition model to obtain a target face gesture to which the target face image belongs; selecting a face recognition model corresponding to the target face posture from the plurality of face recognition models to serve as a target face recognition model; performing feature extraction on the target face image by using the target face recognition model to obtain target face features corresponding to the target face posture; and obtaining the recognition result of the target face image by using the target face characteristic and the reference face characteristic corresponding to the target face posture.
The method for obtaining the recognition result of the target face image by using the target face feature and the reference face feature corresponding to the target face posture comprises the following steps: determining a plurality of reference facial features corresponding to the target facial pose from the reference feature set, and finding out the reference facial features matched with the target facial features from the plurality of reference facial features, wherein the reference feature set comprises the reference facial features corresponding to the plurality of facial poses; and obtaining a recognition result by utilizing the searched reference facial features.
Wherein, prior to determining the plurality of reference facial features corresponding to the target facial pose from the reference feature set, the method further comprises: acquiring reference face images of a plurality of face poses; for each face pose, extracting reference face features of a reference face image of the face pose by using a face recognition model of the face pose; and saving the extracted reference facial features into a reference feature set.
Wherein, obtain the reference facial image of several kinds of facial gestures, include: acquiring input reference face images of a plurality of face poses; or acquiring input reference face images of a first number of face poses, and performing three-dimensional face reconstruction by using the reference face images of the first number of face poses to obtain reference face images of a second number of face poses, wherein the first number of face poses and the second number of face poses are different face poses.
Wherein, from the plurality of reference facial features, finding a reference facial feature matching the target facial feature comprises: and searching out a reference facial feature with the similarity meeting preset requirements with the target facial feature from the plurality of reference facial features.
Wherein, the reference face feature found out is utilized to obtain the recognition result, which comprises: and taking the found reference facial features and/or the identity information corresponding to the found reference facial features as a recognition result.
The method for recognizing the target face image by utilizing the gesture recognition model to obtain the target face gesture to which the target face image belongs comprises the following steps: carrying out gesture recognition on the target face image by using a gesture recognition model to obtain at least one gesture angle; based on the at least one pose angle, a target facial pose to which the target facial image belongs is determined.
Before the posture recognition model is used for carrying out posture recognition on the target face image so as to obtain the target face posture to which the target face image belongs, the method further comprises at least one of the following steps: acquiring first sample face images of a plurality of face postures, and respectively training a face recognition model corresponding to each face posture by using the first sample face image of each face posture; training a pose recognition model using the second sample facial image of the at least one facial pose.
The first sample facial image of at least one facial pose is obtained by carrying out three-dimensional facial reconstruction by using the first sample facial images of other facial poses; and/or the second sample face image of at least one face pose is obtained by performing three-dimensional face reconstruction by using the second sample face images of other face poses.
The face posture comprises a plurality of preset postures of a front face, a left face, a right face, a head-down posture and a head-up posture; and/or the face gesture comprises a plurality of preset gestures and a combined gesture, and the combined gesture consists of more than two preset gestures.
In order to solve the above technical problem, another technical solution adopted by the present application is: there is provided a face recognition device comprising a processor and a memory, the memory for storing program data, the processor for executing the program data to implement the method of any one of the above.
In order to solve the above technical problem, another technical solution adopted by the present application is: there is provided a computer readable storage medium for storing program data, the program data being executable to implement the method of any one of the above.
In the scheme, the target facial image is identified by selecting the facial recognition model according to the facial pose of the object to be identified to obtain the target facial feature, and the target facial feature is compared with the reference facial feature according to the facial pose to obtain the facial recognition result, so that in the facial recognition process, the corresponding matched facial recognition model can be selected according to the target facial pose of the object to be identified to perform feature extraction, the reference facial feature corresponding to the target facial pose is selected to be compared with the extracted target facial feature to realize facial recognition, the feature extraction and feature comparison under the corresponding poses can be performed according to different facial poses to realize the targeted recognition of different facial poses, the limitation on the facial poses of a user during the facial recognition can be reduced, and the accuracy of the facial recognition under different facial poses such as non-frontal poses can be improved, in addition, different recognition models are adopted for different face postures for recognition processing, and compared with the method that a single recognition model corresponds to all face postures, the method can reduce the number of model parameters.
Drawings
FIG. 1 is a schematic flow chart diagram illustrating an embodiment of a face recognition method according to the present application;
FIG. 2 is a schematic flow chart diagram illustrating another embodiment of the face recognition method of the present application;
FIG. 3 is a schematic flow chart diagram illustrating a further embodiment of the face recognition method of the present application;
FIG. 4 is a flowchart illustrating step S310 of a further embodiment of the face recognition method of the present application;
FIG. 5 is a block diagram of an embodiment of the face recognition device of the present application;
FIG. 6 is a block diagram of an embodiment of a computer-readable storage medium of the present application.
Detailed Description
In order to make the purpose, technical solution and effect of the present application clearer and clearer, the present application is further described in detail below with reference to the accompanying drawings and examples.
The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship. Further, the term "plurality" herein means two or more than two.
It is understood that the methods of the present application can include any of the method embodiments described below as well as any non-conflicting combinations of the method embodiments described below.
The face recognition method can be executed by a face recognition device, and the face recognition device can be any device with processing capability, such as a computer, a tablet computer, a mobile phone and the like.
In some embodiments, the face recognition device may perform face recognition on the acquired face image, and the acquisition of the face image may be implemented by an image acquisition module in the face recognition device, or an external image acquisition module connected to the face recognition device, and the image acquisition module may be a camera or the like.
In a specific application, the face recognition device may control the image acquisition module to perform image acquisition, and the specific process may include: the face recognition device provides a collection interface for a user, responds to the operation of the user on the collection interface, starts to collect a face image by using the image collection module, judges whether the collected real-time face image meets the identification requirement, takes the real-time face image meeting the identification requirement as a target face image, and can finish the collection of the face image or continue to obtain the next frame of the target face image. After the face image acquisition is finished, the face recognition device recognizes the target face image by the following face recognition method.
Referring to fig. 1, fig. 1 is a schematic flow chart of an embodiment of a face recognition method according to the present application, which is executed by a face recognition device. In this embodiment, the face recognition method includes:
step S110: and carrying out gesture recognition on the target face image by using the gesture recognition model so as to obtain the target face gesture to which the target face image belongs.
It is to be appreciated that prior to performing gesture recognition, the face recognition device may determine a number of preset gestures, such as multiple of frontal, left, right, head-down, head-up, as recognizable facial gestures, based on user input. Therefore, the gesture recognition model can be obtained through pre-training and used for recognizing the face gesture, then, after the face recognition equipment obtains the target face image acquired by the image acquisition module and acquired by the object to be recognized, the gesture recognition is carried out on the target face image through the gesture recognition model, the target face gesture of the target face image is obtained, and therefore face recognition can be carried out subsequently according to the target face gesture.
In a specific scene, the face recognition device determines preset postures according to user input, wherein the preset postures comprise a front face, a left face, a right face, a head-down posture and a head-up posture, then determines the front face, the left face and the right face as recognizable face postures according to selection of a user in the preset postures, and trains a posture recognition model in advance for recognizing the front face, the left face and the right face. After obtaining the target face image, the face recognition device uses the gesture recognition model to recognize the face gesture of the target face image, and obtains the target face gesture, such as a left face.
In some embodiments, the recognizable facial pose may include a combined pose in addition to the preset pose, wherein the combined pose is composed of two or more preset poses. For example, the preset postures may be a front face, a left face, a right face, a head-down face, a head-up face, and the like, and the combined posture may be a head-down left face, a head-up right face, and the like, so the facial postures recognizable by the posture recognition model obtained by the corresponding training may be a front face, a left face, a head-down right face, and the like.
Step S120: from among the plurality of face recognition models, a face recognition model corresponding to the target face pose is selected as a target face recognition model.
Specifically, several face recognition models are prestored in the face recognition device for performing face recognition by using respectively recognizable face gestures, and for example, the face recognition models may include a left face recognition model, a low head recognition model, a front face recognition model, a high head recognition model, a right face recognition model, and the like. The embodiment can utilize different face recognition models to perform face recognition on face images of different face postures so as to improve the face recognition accuracy rate under different face postures.
In this embodiment, since the face recognition device has different face recognition models corresponding to different face poses, before the face recognition device performs face recognition using the face recognition model, the face recognition device needs to select the corresponding face recognition model according to the target face pose obtained in step S110 to serve as the target face recognition model, and then, the target face image may be subjected to face recognition using the target face recognition model.
Step S130: and performing feature extraction on the target face image by using the target face recognition model to obtain target face features corresponding to the target face posture.
It is understood that the facial features of different objects to be recognized are different, so that the facial recognition device performs facial recognition by extracting facial features in the facial image.
Specifically, for example, the face recognition device performs gesture recognition on the target face image by using the gesture recognition model to obtain a target face gesture to which the target face image belongs as a left face, selects the face recognition model corresponding to the left face as the target face recognition model, and performs feature extraction on the target face image to obtain a target face feature corresponding to the target face image.
Step S140: and obtaining the recognition result of the target face image by using the target face characteristic and the reference face characteristic corresponding to the target face posture.
Specifically, the face recognition device may pre-store reference facial features corresponding to different poses of different objects for use as a basis for face recognition. The different pre-stored postures corresponding to each object may correspond to the postures corresponding to the face recognition model, for example, the postures corresponding to the face recognition model include a front face, a left face and a right face, and then reference facial features under the front face, the left face and the right face of each object are pre-stored respectively. The reference facial features may be acquired by the facial recognition device or obtained externally. After obtaining the target facial features, the face recognition device obtains reference facial features corresponding to the target facial poses of different objects from pre-stored information, and compares the target facial features with the obtained reference facial features respectively, so as to obtain a recognition result, for example, obtain reference facial features matched with the target facial features, and/or obtain identity information of an object to which the target facial features belong, that is, identity information of an object to which the reference facial features matched with the target facial features belong.
In the embodiment, the target facial image is identified by selecting the facial recognition model according to the facial pose of the object to be identified to obtain the target facial feature, and the target facial feature is compared with the reference facial feature according to the facial pose to obtain the facial recognition result, so that in the facial recognition process, the corresponding matched facial recognition model can be selected according to the target facial pose of the object to be identified to extract the feature, and the reference facial feature corresponding to the target facial pose is selected to be compared with the extracted target facial feature to realize the facial recognition, so that the feature extraction and the feature comparison under the corresponding poses can be carried out according to different facial poses to realize the targeted recognition of different facial poses, the limitation on the facial poses of the user during the facial recognition can be reduced, and the accuracy of the facial recognition under different facial poses such as non-frontal poses can be improved, in addition, different recognition models are adopted for different face postures for recognition processing, and compared with the method that a single recognition model corresponds to all face postures, the method can reduce the number of model parameters.
Referring to fig. 2, fig. 2 is a schematic flow chart of another embodiment of the face recognition method of the present application.
It will be appreciated that the face recognition device determines a number of preset poses based on user input and determines a plurality of the preset poses as recognizable facial poses based on user selection. After determining the face pose, the face recognition apparatus may train a face recognition model and a pose recognition model corresponding to the face pose, or acquire the face recognition model and the pose recognition model corresponding to the face pose from the outside. In the present embodiment, the face recognition device trains the face recognition model and the posture recognition model as an example. The face recognition method comprises the following steps:
step S210: and acquiring first sample face images of a plurality of face postures, and training face recognition models corresponding to each face posture by using the first sample face images of each face posture.
Specifically, after determining a plurality of recognizable facial poses, the face recognition device acquires a first sample face image for each facial pose, that is, acquires first sample face images labeled as each facial pose respectively, for example, if the recognizable facial poses include a front face, a left face and a right face, at least one first sample face image labeled as the front face, at least one first sample face image labeled as the left face, and at least one first sample face image labeled as the right face are acquired respectively. The first sample face image may be pre-stored by the face recognition device, or may be obtained by the face recognition device from the outside. Specifically, the labeling postures of the first sample facial image may all be artificially labeled, or part of the first sample facial image is artificially labeled, and the rest of the first sample facial image is obtained by the face recognition device by using semi-supervised learning labeling.
In some embodiments, due to reasons such as limited acquisition conditions, the face recognition device cannot directly acquire the first sample face image of each face pose, the face recognition device may acquire the first sample face image corresponding to some face poses acquired by the image acquisition module for an object, perform three-dimensional face reconstruction using the acquired first sample face image, and then obtain the first sample face image corresponding to other face poses of the object based on the three-dimensional reconstructed face model. Specifically, the face recognition device performs three-dimensional face reconstruction by using the face information contained in the acquired first sample face image, so as to acquire corresponding first sample face images in other postures, the more the acquired face information is, the more the acquired first sample face image is accurate, and the more the face information contained in the front face is, so that the acquisition of the first sample face image containing the front face is more beneficial to acquiring the accurate first sample face image in other postures.
The face recognition equipment trains a face recognition model of each face gesture by using the acquired first sample face image corresponding to each face gesture, so that the face recognition model of each face gesture performs face recognition on the face image of the corresponding face gesture, and then the output face features meet the preset face recognition requirements.
Step S220: training a pose recognition model using the second sample facial image of the at least one facial pose.
Specifically, after determining a plurality of recognizable facial poses, the face recognition device determines a reference pose angle range corresponding to each facial pose, and acquires a second sample face image labeled with a labeled pose angle for all or part of all the facial poses, for example, if the recognizable facial poses include a front face and a left face, a first face image labeled with a labeled pose angle is acquired. Specifically, the labeling attitude angles of the second sample face image may all be artificially labeled, or part of the second sample face image is artificially labeled, and the rest of the second sample face image is obtained by the face recognition device by using semi-supervised learning labeling. Wherein, part or all of the second sample face image can be the same as the first sample face image, and the second sample image can be pre-stored in the face recognition device or obtained from the outside.
In some embodiments, the face recognition device may obtain a second sample face image of some face poses acquired by the image acquisition module for an object, perform three-dimensional face reconstruction by using the acquired second sample face image, and further obtain a second sample face image corresponding to other face poses of the object based on the three-dimensional reconstructed face model. The process of generating the second sample face image corresponding to the other face pose by performing the three-dimensional face reconstruction is similar to the process of generating the first sample face image by performing the three-dimensional face reconstruction, and is not repeated herein.
The process of performing gesture recognition by the face recognition device specifically includes: the face recognition device performs pose recognition on the face image by using the pose recognition model to obtain a pose angle, and compares the pose angle with a reference pose angle range of each face pose to determine the face pose of the face image. In the process of training the gesture recognition model, the face recognition device compares the gesture angle obtained by recognizing the second sample face image with the labeled gesture angle of the second sample face image, so that the recognition accuracy of the gesture recognition model can be determined. The face recognition device can gradually adjust the parameters of the gesture recognition model, improve the recognition accuracy of the gesture recognition model, and finish the training of the gesture recognition model after judging that the recognition accuracy meets the preset gesture recognition requirement.
It is understood that the order of steps S210 and S220 may be exchanged without affecting the training of the gesture recognition model and the face recognition model.
It can be understood that, under the condition that enough sample face images can be obtained, the sample face images can be obtained without adopting a three-dimensional face reconstruction mode, under the condition that the sample face images are limited, by the mode, the face recognition device can utilize limited part of the sample face images to obtain enough sample images through the three-dimensional face reconstruction so as to train the gesture recognition model and the face recognition model, and under the condition that the sample images are limited, the accuracy of the models is improved.
Step S230: and carrying out gesture recognition on the target face image by using the gesture recognition model to obtain at least one gesture angle.
In this embodiment, the pose information output by the pose recognition model is at least one pose angle, for example, an euler angle, and in some specific application scenarios, the pose information includes pose angles in three different directions, for example, pose angles in an X-axis, a Y-axis, and a Z-axis, for example, a face recognition device performs pose recognition on a target face image by using the pose recognition model to obtain three pose angles for determining a target face pose to which the target face image belongs. It is understood that the gesture information output by the gesture recognition model may not be three gesture angles, but may be other number such as one or two or even more than three gesture angles, and the specific number may be determined according to the specific recognizable face gesture, for example, when the recognizable face gesture only includes a frontal face and a head-up, the gesture information output by the gesture recognition model may be only one gesture angle, such as a gesture angle on the Z-axis.
Step S240: based on the at least one pose angle, a target facial pose to which the target facial image belongs is determined.
Specifically, after obtaining the attitude angle of the target face image, the face recognition device compares the attitude angle with a reference attitude angle range corresponding to each face attitude, and determines a reference attitude angle range to which the attitude angle belongs, thereby determining the face attitude corresponding to the reference attitude angle range as the target face attitude.
For example, the recognizable facial pose includes a face-up and a head-up, and after the pose angle of the target facial image is obtained, the pose angle is compared with the reference pose angle ranges of the face-up and the head-up, and the pose angle is determined to belong to the reference pose angle range of the head-up, so that the target facial pose of the target facial image is determined to be the head-up.
It is understood that the above step S110 can be implemented by the step S230 and the step S240.
Step S250: from among the plurality of face recognition models, a face recognition model corresponding to the target face pose is selected as a target face recognition model.
Step S260: and performing feature extraction on the target face image by using the target face recognition model to obtain target face features corresponding to the target face posture.
Step S270: and obtaining the recognition result of the target face image by using the target face characteristic and the reference face characteristic corresponding to the target face posture.
For the detailed description of step S250, step S260 and step S270, reference may be made to the above-mentioned related descriptions about steps S120, S130 and S140, which are not repeated herein.
In the embodiment, the gesture recognition model and the face recognition model are trained according to the recognizable face gesture, the face recognition model is selected according to the face gesture of the object to be recognized to recognize the target face image to obtain the target face feature, and the target face feature is compared with the reference face feature according to the face gesture to obtain the face recognition result, so that in the face recognition process, the corresponding matched face recognition model can be selected according to the target face gesture of the object to be recognized to perform feature extraction, the reference face feature corresponding to the target face gesture is selected to be compared with the extracted target face feature to realize face recognition, so that the feature extraction and feature comparison under the corresponding gestures can be performed according to different face gestures to realize the targeted recognition of different face gestures, and the limitation on the face gesture of the user during the face recognition can be reduced, the accuracy rate of face recognition under different face postures such as non-frontal face posture is improved, in addition, different recognition models are adopted for different face postures for recognition processing, and compared with the situation that a single recognition model corresponds to all face postures, the number of model parameters can be reduced.
Referring to fig. 3, fig. 3 is a schematic flow chart of a further embodiment of the face recognition method of the present application.
It can be understood that, before the user performs the first face recognition, the face recognition device may acquire a reference face image corresponding to the recognizable face pose of the user as a basis for face recognition, and the face recognition device may store the acquired reference face image in a reference image set for subsequent use. The face recognition device may acquire a face image of the user as a reference face image, or may acquire the reference face image from the outside. Therefore, the face recognition method comprises the following steps:
step S310: reference facial images are acquired for a number of facial poses.
Specifically, the face recognition device performs face recognition by performing face recognition on different face poses, so that a corresponding reference face image is required for each recognizable face pose of different objects as a basis. Thus, after determining the recognizable facial poses, the face recognition device acquires a reference face image corresponding to each facial pose of the recognition object, the reference face image being labeled with the corresponding facial pose. For example, the recognizable face pose includes a front face, a left face, and a right face, and the face recognition device acquires a reference face image labeled with the front face, a reference face image labeled with the left face, and a reference face image labeled with the right face, respectively.
The process of face recognition by the face recognition device is performed based on the reference face feature, so that after acquiring the reference face image, the face recognition device acquires the reference face feature using the reference face image for serving as a basis for face recognition.
Step S320: for each face pose, extracting a reference facial feature of a reference facial image of the face pose using a face recognition model of the face pose.
Specifically, the face recognition device acquires a reference face image marked with a face pose, selects a face recognition model according to the face pose, extracts a reference face feature of the reference face image by using the face recognition model, and corresponds the reference face feature with the face pose for subsequent face recognition.
Step S330: and saving the extracted reference facial features into a reference feature set.
Specifically, the face recognition device stores the reference face features and the corresponding face poses into a reference feature set, so that the face recognition device selects a plurality of reference face features corresponding to the target face poses for comparison after performing pose recognition.
Step S340: and carrying out gesture recognition on the target face image by using the gesture recognition model so as to obtain the target face gesture to which the target face image belongs.
Step S350: from among the plurality of face recognition models, a face recognition model corresponding to the target face pose is selected as a target face recognition model.
Step S360: and performing feature extraction on the target face image by using the target face recognition model to obtain target face features corresponding to the target face posture.
For the detailed description of step S340, step S350 and step S360, reference may be made to the detailed description of steps S110, S120 and S130, which is not repeated herein.
Step S370: a plurality of reference facial features corresponding to the target facial pose are determined from the set of reference features, and a reference facial feature matching the target facial feature is found from the plurality of reference facial features.
Specifically, the reference feature set is a set of reference facial features corresponding to each facial pose, after the target facial features and the target facial pose are obtained, the face recognition device determines a plurality of reference facial features corresponding to the target facial pose from the reference feature set, compares the target facial features with the plurality of reference facial features, and finds out, from the plurality of reference facial features, the reference facial features whose similarity to the target facial features meets a preset requirement for obtaining a recognition result, where the preset requirement may be that the similarity reaches a specific value.
For example, the face recognition device obtains the target facial pose as a left face, determines reference facial features of the plurality of left faces from the reference feature set, and finds reference facial features matching the target facial features from the reference facial features of the plurality of left faces.
Step S380: and obtaining a recognition result by utilizing the searched reference facial features.
Specifically, the face recognition device may use at least one of the found reference facial feature, a facial image corresponding to the found reference facial feature, and corresponding identity information as a recognition result, and may display the recognition result for the user to view.
Referring to fig. 3 and 4, fig. 4 is a schematic flow chart of step S310 in a further embodiment of the face recognition method of the present application, and in the present embodiment, a face recognition device is taken as an example to collect a reference face image, and it is understood that the reference face image obtained and input may also be obtained from the outside. The step S310 may specifically include the following steps:
step S411: reference face images of the input several facial poses are acquired.
Before the user carries out the first face recognition, the face recognition device carries out the acquisition of a corresponding reference face image according to the recognizable face gesture so as to obtain the reference face characteristic. Specifically, the face recognition device can provide a collection interface for a user, display a face pose required to be collected on the collection interface so that the user can conveniently make a corresponding face pose, collect a face image by using the image collection module, determine a face pose corresponding to the face image to be collected or already collected in response to user selection, correspondingly store the collected face image and the face pose, and use the collected face image as a reference face image after judging that the collected face image meets the requirements of the reference face image.
It is understood that the manner adopted in step S411 is to directly acquire the reference face image corresponding to each recognizable face pose. In some embodiments, it is not convenient to directly acquire the reference face image corresponding to each recognizable face pose due to limited acquisition conditions, etc., and reference face images of all face poses may be obtained by acquiring reference face images of partial face poses, which is adopted in steps S412 and S413.
It is understood that step S411 and steps S412 and S413 are for acquiring reference face images of several face poses, so the face recognition device can perform step S411 or step S412 and step S413.
Step S412: an input reference facial image of a first number of facial poses is acquired.
The first number is smaller than the total number of the recognizable face gestures, the face recognition device collects reference face images of the recognizable face gestures, and the collected reference face images are used for obtaining reference face images of other gestures, so that a complete reference face image is obtained.
Specifically, the face recognition device obtains other reference face images according to the face information contained in the collected reference face image, the more the collected face information is, the more the face recognition device can obtain other reference face images, the more accurate the obtained reference face image can be, and the more the face information contained in the face image of the front face is, so that the more the collected reference face image containing the face image of the front face is, the more the accurate reference face image can be obtained.
Step S413: three-dimensional face reconstruction is performed using the reference face images of the first number of facial poses to obtain reference face images of a second number of facial poses.
Wherein the first number and the second number are added to form a total number of recognizable facial poses, the first number of facial poses and the second number of facial poses being different types of facial poses. After acquiring the reference face images of the first number of face poses, the face recognition device performs three-dimensional face reconstruction by using face information contained in the acquired reference face images, and then obtains reference face images corresponding to the rest face poses based on a three-dimensional reconstruction face model.
For example, the recognizable face pose includes a front face, a left face, and a right face, and the face recognition device acquires a reference face image of the front face, and performs three-dimensional face reconstruction using the reference face image of the front face, thereby obtaining a reference face image of the left face and a reference face image of the right face.
In the case that reference face images corresponding to all recognizable face poses can be acquired, the step S411 is preferentially executed, in the case that reference face images corresponding to only part of recognizable face poses can be acquired, the step S412 and the step S413 are executed, by this way, the face recognition device can use the limited reference face images to perform three-dimensional face reconstruction so as to obtain the reference face images corresponding to the recognizable face poses for subsequent face recognition, the step of acquiring the reference face images can be simplified, and in the case that the acquired reference face images are limited, different face poses can still be respectively recognized, so that the applicable scenes of the face recognition device are wider.
In the embodiment, the corresponding reference facial image and the reference facial feature are obtained according to the recognizable facial pose, the target facial image is identified by selecting the facial identification model according to the facial pose of the object to be identified to obtain the target facial feature, and the reference facial feature corresponding to the target facial feature is compared according to the facial pose to obtain the facial identification result, so that in the facial identification process, the corresponding matched facial identification model can be selected according to the target facial pose of the object to be identified for feature extraction, the reference facial feature corresponding to the target facial pose is selected for comparison with the extracted target facial feature to realize facial identification, so that the feature extraction and feature comparison under the corresponding pose can be carried out according to different facial poses to realize the targeted identification of different facial poses, and the limitation on the facial pose of the user during the facial identification can be reduced, the accuracy rate of face recognition under different face postures such as non-frontal face posture is improved, in addition, different recognition models are adopted for different face postures for recognition processing, and compared with the situation that a single recognition model corresponds to all face postures, the number of model parameters can be reduced.
Referring to fig. 5, fig. 5 is a schematic diagram of a framework of an embodiment of the face recognition device of the present application.
In the present embodiment, the face recognition device 50 includes a memory 51 and a processor 52, wherein the memory 51 is connected to the processor 52. In some embodiments, the face recognition device 50 may further include a human-computer interaction circuit 53 and an image acquisition module 54, the memory 51, the human-computer interaction circuit 53 and the image acquisition module 54 are respectively coupled to the processor 52, specifically, the respective components of the face recognition device 50 may be coupled together through a bus, or the processor 52 of the face recognition device 50 is respectively connected with the other components one by one.
The face recognition device 50 may be the face recognition device in any of the above embodiments, and may specifically be any device with processing capability, such as a computer, a tablet computer, a mobile phone, and the like.
The memory 51 is used for storing program data executed by the processor 52 and data of the processor 52 in the process. Such as a pose recognition model, a face recognition model, a set of reference features, a target face image, a target facial feature, and so forth. The memory 51 includes a nonvolatile storage portion for storing the program data.
The processor 52 controls the operation of the face recognition device 50, and the processor 52 may also be referred to as a CPU (Central Processing Unit). Processor 52 may be an integrated circuit chip having signal processing capabilities. The processor 52 may also be a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. In addition, the processor 52 may be commonly implemented by a plurality of circuit-forming chips.
The man-machine interaction circuit 53 is used for displaying the acquisition interface and the identification result and receiving the operation of the user on the acquisition interface.
The image acquisition module 54 is used for acquiring a user face image for face recognition, and the face image may include a reference face image, a target face image, and the like.
In the embodiment, the target facial image is identified by selecting the facial recognition model according to the facial pose of the object to be identified to obtain the target facial feature, and the target facial feature is compared with the reference facial feature according to the facial pose to obtain the facial recognition result, so that in the facial recognition process, the corresponding matched facial recognition model can be selected according to the target facial pose of the object to be identified to extract the feature, and the reference facial feature corresponding to the target facial pose is selected to be compared with the extracted target facial feature to realize the facial recognition, so that the feature extraction and the feature comparison under the corresponding poses can be carried out according to different facial poses to realize the targeted recognition of different facial poses, the limitation on the facial poses of the user during the facial recognition can be reduced, and the accuracy of the facial recognition under different facial poses such as non-frontal poses can be improved, in addition, different recognition models are adopted for different face postures for recognition processing, and compared with the method that a single recognition model corresponds to all face postures, the method can reduce the number of model parameters.
Referring to fig. 6, fig. 6 is a block diagram of an embodiment of a computer-readable storage medium according to the present application.
In this embodiment, the computer readable storage medium 60 stores processor executable program data 61 that can be executed to implement any of the methods described above.
The computer-readable storage medium 60 may be a medium that can store program data, such as a usb disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, or may be a server that stores the program data, and the server may send the stored program data to another device for operation, or may self-operate the stored program data.
In some embodiments, computer-readable storage medium 60 may also be a memory as shown in FIG. 5.
In the embodiment, the target facial image is identified by selecting the facial recognition model according to the facial pose of the object to be identified to obtain the target facial feature, and the target facial feature is compared with the reference facial feature according to the facial pose to obtain the facial recognition result, so that in the facial recognition process, the corresponding matched facial recognition model can be selected according to the target facial pose of the object to be identified to extract the feature, and the reference facial feature corresponding to the target facial pose is selected to be compared with the extracted target facial feature to realize the facial recognition, so that the feature extraction and the feature comparison under the corresponding poses can be carried out according to different facial poses to realize the targeted recognition of different facial poses, the limitation on the facial poses of the user during the facial recognition can be reduced, and the accuracy of the facial recognition under different facial poses such as non-frontal poses can be improved, in addition, different recognition models are adopted for different face postures for recognition processing, and compared with the method that a single recognition model corresponds to all face postures, the method can reduce the number of model parameters.
The above description is only for the purpose of illustrating embodiments of the present application and is not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application or are directly or indirectly applied to other related technical fields, are also included in the scope of the present application.
Claims (12)
1. A method of face recognition, the method comprising:
carrying out gesture recognition on the target face image by using a gesture recognition model to obtain a target face gesture to which the target face image belongs;
selecting a face recognition model corresponding to the target face posture from a plurality of face recognition models to serve as a target face recognition model;
performing feature extraction on the target face image by using the target face recognition model to obtain target face features corresponding to the target face posture;
and obtaining the recognition result of the target face image by using the target face feature and the reference face feature corresponding to the target face posture.
2. The method according to claim 1, wherein the obtaining a recognition result of the target facial image by using the target facial feature and a reference facial feature corresponding to the target facial pose comprises:
determining a plurality of reference facial features corresponding to the target facial pose from a reference feature set, and finding a reference facial feature matching the target facial feature from the plurality of reference facial features, wherein the reference feature set comprises reference facial features corresponding to a plurality of facial poses;
and obtaining the identification result by utilizing the searched reference facial features.
3. The method of claim 2, wherein prior to said determining a plurality of reference facial features corresponding to the target facial pose from a reference feature set, the method further comprises:
acquiring reference face images of a plurality of face poses;
for each face pose, extracting a reference facial feature of a reference facial image of the face pose by using a face recognition model of the face pose;
and saving the extracted reference facial features into the reference feature set.
4. The method of claim 3, wherein said obtaining reference facial images for a plurality of said facial poses comprises:
acquiring input reference face images of a plurality of face poses; alternatively, the first and second electrodes may be,
the method comprises the steps of obtaining input reference facial images of a first number of facial poses, and conducting three-dimensional facial reconstruction by utilizing the reference facial images of the first number of facial poses to obtain reference facial images of a second number of facial poses, wherein the first number of facial poses and the second number of facial poses are different facial poses.
5. The method of claim 2, wherein the finding out the reference facial feature matching the target facial feature from the plurality of reference facial features comprises:
and searching out a reference facial feature with the similarity meeting preset requirements with the target facial feature from the plurality of reference facial features.
6. The method according to claim 2, wherein the obtaining the recognition result by using the found reference facial feature comprises:
and taking the found reference facial features and/or the identity information corresponding to the found reference facial features as the identification result.
7. The method of claim 1, wherein the performing gesture recognition on the target facial image by using a gesture recognition model to obtain a target facial gesture to which the target facial image belongs comprises:
carrying out gesture recognition on the target face image by using a gesture recognition model to obtain at least one gesture angle;
based on the at least one pose angle, a target facial pose to which the target facial image belongs is determined.
8. The method of claim 1, wherein before the gesture recognition of the target facial image using the gesture recognition model to obtain the target facial gesture to which the target facial image belongs, the method further comprises at least one of:
acquiring first sample face images of a plurality of face postures, and respectively training a face recognition model corresponding to each face posture by using the first sample face image of each face posture;
training the pose recognition model using a second sample facial image of at least one of the facial poses.
9. The method of claim 8, wherein said first sample facial image of at least one of said facial poses is obtained by three-dimensional facial reconstruction using first sample facial images of other of said facial poses;
and/or the second sample facial image of at least one face pose is obtained by performing three-dimensional facial reconstruction by using second sample facial images of other face poses.
10. The method of claim 1, wherein the facial gestures include a plurality of preset gestures of face front, left face, right face, head down, head up;
and/or the face gesture comprises a plurality of preset gestures and a combined gesture, and the combined gesture is composed of more than two preset gestures.
11. A face recognition device, characterized in that the face recognition device comprises a processor and a memory for storing program data, the processor being adapted to execute the program data to implement the method according to any of claims 1-10.
12. A computer-readable storage medium for storing program data executable to implement the method of any one of claims 1-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110845152.1A CN113657187A (en) | 2021-07-26 | 2021-07-26 | Face recognition method, face recognition equipment and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110845152.1A CN113657187A (en) | 2021-07-26 | 2021-07-26 | Face recognition method, face recognition equipment and computer-readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113657187A true CN113657187A (en) | 2021-11-16 |
Family
ID=78490253
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110845152.1A Pending CN113657187A (en) | 2021-07-26 | 2021-07-26 | Face recognition method, face recognition equipment and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113657187A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114387651A (en) * | 2022-01-12 | 2022-04-22 | 北京百度网讯科技有限公司 | Face recognition method, device, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304829A (en) * | 2018-03-08 | 2018-07-20 | 北京旷视科技有限公司 | Face identification method, apparatus and system |
| CN109726648A (en) * | 2018-12-14 | 2019-05-07 | 深圳壹账通智能科技有限公司 | A kind of facial image recognition method and device based on machine learning |
| CN110490067A (en) * | 2019-07-11 | 2019-11-22 | 平安科技（深圳）有限公司 | A kind of face identification method and device based on human face posture |
| CN110647865A (en) * | 2019-09-30 | 2020-01-03 | 腾讯科技（深圳）有限公司 | Face gesture recognition method, device, equipment and storage medium |
| CN111414803A (en) * | 2020-02-24 | 2020-07-14 | 北京三快在线科技有限公司 | Face recognition method and device and electronic equipment |
-
2021
- 2021-07-26 CN CN202110845152.1A patent/CN113657187A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304829A (en) * | 2018-03-08 | 2018-07-20 | 北京旷视科技有限公司 | Face identification method, apparatus and system |
| CN109726648A (en) * | 2018-12-14 | 2019-05-07 | 深圳壹账通智能科技有限公司 | A kind of facial image recognition method and device based on machine learning |
| CN110490067A (en) * | 2019-07-11 | 2019-11-22 | 平安科技（深圳）有限公司 | A kind of face identification method and device based on human face posture |
| CN110647865A (en) * | 2019-09-30 | 2020-01-03 | 腾讯科技（深圳）有限公司 | Face gesture recognition method, device, equipment and storage medium |
| CN111414803A (en) * | 2020-02-24 | 2020-07-14 | 北京三快在线科技有限公司 | Face recognition method and device and electronic equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114387651A (en) * | 2022-01-12 | 2022-04-22 | 北京百度网讯科技有限公司 | Face recognition method, device, equipment and storage medium |
| CN114387651B (en) * | 2022-01-12 | 2023-10-20 | 北京百度网讯科技有限公司 | Face recognition method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10445562B2 (en) | AU feature recognition method and device, and storage medium | |
| US10534957B2 (en) | Eyeball movement analysis method and device, and storage medium | |
| US9367756B2 (en) | Selection of representative images | |
| CN111626371B (en) | Image classification method, device, equipment and readable storage medium | |
| WO2020078017A1 (en) | Method and apparatus for recognizing handwriting in air, and device and computer-readable storage medium | |
| WO2019033573A1 (en) | Facial emotion identification method, apparatus and storage medium | |
| CN107483834B (en) | Image processing method, continuous shooting method and device and related medium product | |
| JP5366756B2 (en) | Information processing apparatus and information processing method | |
| JP5361524B2 (en) | Pattern recognition system and pattern recognition method | |
| US8452091B2 (en) | Method and apparatus for converting skin color of image | |
| CN110741377A (en) | Face image processing method and device, storage medium and electronic equipment | |
| Vazquez-Fernandez et al. | Built-in face recognition for smart photo sharing in mobile devices | |
| CN110941992B (en) | Smile expression detection method and device, computer equipment and storage medium | |
| JP7107598B2 (en) | Authentication face image candidate determination device, authentication face image candidate determination method, program, and recording medium | |
| US20180107877A1 (en) | Image processing apparatus, image processing method, and image processing system | |
| WO2023284182A1 (en) | Training method for recognizing moving target, method and device for recognizing moving target | |
| CN112241667A (en) | Image detection method, device, equipment and storage medium | |
| CN112417970A (en) | Target object identification method, device and electronic system | |
| EP3200092A1 (en) | Method and terminal for implementing image sequencing | |
| CN113657187A (en) | Face recognition method, face recognition equipment and computer-readable storage medium | |
| CN109697421A (en) | Evaluation method, device, computer equipment and storage medium based on micro- expression | |
| JP2021026744A (en) | Information processing device, image recognition method, and learning model generation method | |
| CN105512664A (en) | Image recognition method and device | |
| JP5286574B2 (en) | Object detection recognition apparatus, object detection recognition method, and program | |
| CN108596127A (en) | A kind of fingerprint identification method, auth method and device and identity veritify machine |
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 |