CN110188630A - A kind of face identification method and camera - Google Patents
A kind of face identification method and camera Download PDFInfo
- Publication number
- CN110188630A CN110188630A CN201910395391.4A CN201910395391A CN110188630A CN 110188630 A CN110188630 A CN 110188630A CN 201910395391 A CN201910395391 A CN 201910395391A CN 110188630 A CN110188630 A CN 110188630A
- Authority
- CN
- China
- Prior art keywords
- face
- facial
- angle
- image
- facial 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of face identification method and camera.The method comprise the steps that obtaining the face feature point of facial image;Facial angle, face width are obtained according to the face feature point, and according to the eigenmatrix of multiple source images in the feature vector and image library being made of the facial features, obtains the human face similarity degree between the facial image and the multiple source images;Recognition of face is carried out according to the facial angle, the face width and the human face similarity degree, obtains the recognition result of the facial image.Technical solution of the present invention is based on much information and identifies to facial image, can be improved the accuracy of recognition of face, it is ensured that face can be accurately and rapidly identified in most scenes.
Description
Technical field
The present invention relates to machine learning techniques field more particularly to a kind of face identification methods and camera.
Background technique
Currently, the gradually mature and computer hardware performance high speed with machine learning techniques is promoted, computer in recent years
The application fields such as vision, natural language processing and speech recognition achieve breakthrough.Recognition of face is as computer vision
The basic task in one, field, precision are also significantly enhanced.The development of depth learning technology solves recognition of face and works as
In feature representation the problem of, compared to traditional method, the feature of face can be learnt more fully hereinafter.
But under practice scene, there is also the problems of recognition of face difficulty for existing face identification method.
Summary of the invention
The present invention provides a kind of face identification method and cameras, difficult to solve existing face identification method recognition of face
The problem of.
In a first aspect, the present invention provides a kind of face identification methods, comprising: obtain the face feature point of facial image;
Facial angle, face width are obtained according to the face feature point, and according to the feature vector being made of the facial features
With the eigenmatrix of source images multiple in image library, the face obtained between the facial image and the multiple source images is similar
Degree;Recognition of face is carried out according to the facial angle, the face width and the human face similarity degree, obtains the facial image
Recognition result.
In some embodiments, face is carried out according to the facial angle, the face width and the human face similarity degree
Identification, obtains the recognition result of the facial image, comprising: the facial angle is compared with predetermined angle threshold value,
When the facial angle is less than the predetermined angle threshold value, identify that the facial image is invalid image, in the facial angle
When not less than the predetermined angle threshold value, the face width is compared with predetermined width threshold value, in the face width
When less than the predetermined width threshold value, identify that the facial image is invalid image, in the face width not less than described pre-
If when width threshold value, obtaining the relationship of the human face similarity degree and the first similarity threshold and the second similarity threshold, described
One similarity threshold is greater than second similarity threshold;According to the human face similarity degree and first similarity threshold and institute
The relationship for stating the second similarity threshold obtains the recognition result of facial image.
In some embodiments, according to the human face similarity degree and first similarity threshold and second similarity
The relationship of threshold value obtains the recognition result of facial image, comprising: if the human face similarity degree is greater than the first similarity threshold
Value identifies that the corresponding user of the facial image is target source images owning user in described image library, the target source images
For the corresponding source images of human face similarity degree maximum value in the multiple source images between the facial image;If the face
Similarity is less than second similarity threshold, identifies that the corresponding user of the facial image is new user, obtains the new use
The information of the facial image and the new user is simultaneously added in described image library by the information at family;If the human face similarity degree
No more than first similarity threshold and the human face similarity degree is not less than second similarity threshold, reacquires face
The face feature point of image.
In some embodiments, the face feature point of facial image is obtained, comprising: utilize multitask concatenated convolutional network pair
The facial image carries out Face datection, alternatively, Face datection is carried out to the facial image using Machine learning tools Dlib,
Obtain the human face region subgraph of the facial image;The human face region subgraph is input to trained convolutional Neural net
Network obtains the face feature point of the facial image.
In some embodiments, facial angle, face width are obtained according to the face feature point, comprising: according to acquisition
The camera internal parameter of the facial image obtains match point of the face feature point in camera coordinates system, the matching
Plane of the point where in the camera coordinates system is benchmark plane;3D standard faces mould is constructed in the camera coordinates system
Type includes multiple three-dimensional facial features points in the 3D standard faces model;In the camera coordinates system, according to rotation peace
Multiple mapping points that the multiple three-dimensional facial features point is formed on the datum plane after shifting processing and the match point
It is overlapped the most rotation angle of quantity, obtains the facial angle;According to both sides of the edge facial characteristics in the face feature point
The distance between point, obtains the face width.
In some embodiments, rotation and translation processing includes: to the multiple three-dimensional facial features point with respect to XOY plane
It is rotated, the multiple three-dimensional facial features point rotate and to the multiple three-dimensional along the direction parallel with Y-axis
Face feature point is rotated along the direction parallel with X-axis;Wherein, OXYZ is the camera coordinates system, and O is camera seat
The origin of system is marked, X, Y and Z are respectively three reference axis of camera coordinates system, and the XOY plane is the datum plane.
In some embodiments, the facial angle includes: swing angle, rotational angle and pitch angle;The swing
Angle corresponds to the multiple three-dimensional facial features point relative to the rotation angle of XOY plane, and the rotational angle corresponds to described
Multiple three-dimensional facial features points correspond to the multiple three-dimensional along the rotation angle with Y-axis parallel direction, the pitch angle
Face feature point is along the rotation angle with X-axis parallel direction.
In some embodiments, according to multiple source images in the feature vector and image library being made of the facial features
Eigenmatrix obtains the human face similarity degree between the facial image and the multiple source images, comprising: calculates separately the people
Cosine similarity in the feature vector of face image and the eigenmatrix between the feature vector of each source images, calculating is arrived
The cosine similarity in maximum value be determined as the human face similarity degree between the facial image and the multiple source images.
Second aspect, the present invention provides a kind of cameras, comprising: camera and processor;The camera acquires face
Image is simultaneously sent to the processor;The processor obtains the face feature point of facial image;According to the face feature point
Facial angle, face width are obtained, and according to source figures multiple in the feature vector and image library being made of the facial features
The eigenmatrix of picture obtains the human face similarity degree between the facial image and the multiple source images;According to the face angle
Degree, the face width and the human face similarity degree carry out recognition of face, obtain the recognition result of the facial image.
In some embodiments, the facial angle is compared, described by the processor with predetermined angle threshold value
When facial angle is less than the predetermined angle threshold value, identify that the facial image is invalid image, it is not small in the facial angle
When the predetermined angle threshold value, the face width is compared with predetermined width threshold value, is less than in the face width
It when the predetermined width threshold value, identifies that the facial image is invalid image, is not less than the default width in the face width
When spending threshold value, the relationship of the human face similarity degree and the first similarity threshold and the second similarity threshold, first phase are obtained
It is greater than second similarity threshold like degree threshold value;If the human face similarity degree is greater than first similarity threshold, institute is identified
Stating the corresponding user of facial image is target source images owning user in described image library, and the target source images are the multiple
The corresponding source images of human face similarity degree maximum value in source images between the facial image;If the human face similarity degree is less than
Second similarity threshold identifies that the corresponding user of the facial image is new user, obtains the information of the new user simultaneously
The information of the facial image and the new user is added in described image library;If the human face similarity degree is no more than described
First similarity threshold and the human face similarity degree are not less than second similarity threshold, reacquire the face of facial image
Characteristic point.
The present invention obtains face feature point by identifying to facial image, obtains facial angle, people based on face feature point
Human face similarity degree information between face width and facial image and source images knows facial image based on much information
Not, the accuracy of recognition of face is improved, it is ensured that face can be accurately and rapidly identified in most scenes.
Detailed description of the invention
Fig. 1 is the flow chart of the face identification method shown in the embodiment of the present invention;
Fig. 2 is the face recognition process schematic diagram shown in the embodiment of the present invention;
Fig. 3 is the facial angle schematic diagram shown in the embodiment of the present invention;
Fig. 4 is the structural block diagram of the camera shown in the embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Hereinafter, will be described with reference to the accompanying drawings the embodiment of the present invention.However, it should be understood that these descriptions are only exemplary
, and be not intended to limit the scope of the invention.In addition, in the following description, descriptions of well-known structures and technologies are omitted, with
Avoid unnecessarily obscuring idea of the invention.
Term as used herein is not intended to limit the present invention just for the sake of description specific embodiment.Used here as
Word " one ", " one (kind) " and "the" etc. also should include " multiple ", " a variety of " the meaning, unless in addition context clearly refers to
Out.In addition, the terms "include", "comprise" as used herein etc. show the presence of the feature, step, operation and/or component,
But it is not excluded that in the presence of or add other one or more features, step, operation or component.
There are all terms (including technical and scientific term) as used herein those skilled in the art to be generally understood
Meaning, unless otherwise defined.It should be noted that term used herein should be interpreted that with consistent with the context of this specification
Meaning, without that should be explained with idealization or excessively mechanical mode.
Shown in the drawings of some block diagrams and/or flow chart.It should be understood that some sides in block diagram and/or flow chart
Frame or combinations thereof can be realized by computer program instructions.These computer program instructions can be supplied to general purpose computer,
The processor of special purpose computer or other programmable data processing units, so that these instructions are when executed by this processor can be with
Creation is for realizing function/operation device illustrated in these block diagrams and/or flow chart.
Therefore, technology of the invention can be realized in the form of hardware and/or software (including firmware, microcode etc.).Separately
Outside, technology of the invention can take the form of the computer program product on the machine readable media for being stored with instruction, the meter
Calculation machine program product uses for instruction execution system or instruction execution system is combined to use.In the context of the present invention,
Machine readable media, which can be, can include, store, transmitting, propagating or transmitting the arbitrary medium of instruction.For example, machine readable Jie
Matter can include but is not limited to electricity, magnetic, optical, electromagnetic, infrared or semiconductor system, device, device or propagation medium.It is machine readable
The specific example of medium includes: magnetic memory apparatus, such as tape or hard disk (HDD);Light storage device, such as CD (CD-ROM);It deposits
Reservoir, such as random access memory (RAM) or flash memory;And/or wire/wireless communication link.
The present invention provides a kind of face identification method.
Fig. 1 is the flow chart of the face identification method shown in the embodiment of the present invention, as shown in Figure 1, the method for the present embodiment
Include:
S110 obtains the face feature point of facial image.
Wherein, face feature point includes the multiple features for identifying the positions such as eyes, nose, mouth, eyebrow, facial contour
Point.
S120 obtains facial angle, face width according to the face feature point, and according to by the facial features group
At feature vector and image library in multiple source images eigenmatrix, obtain the facial image and the multiple source images it
Between human face similarity degree.
S130 carries out recognition of face according to the facial angle, the face width and the human face similarity degree, obtains institute
State the recognition result of facial image.
The present embodiment by facial image identify obtain face feature point, based on face feature point obtain facial angle,
Human face similarity degree information between face width and facial image and source images knows facial image based on much information
Not, the accuracy of recognition of face is improved, it is ensured that face can be accurately and rapidly identified in most scenes.
Fig. 2 is the face recognition process schematic diagram shown in the embodiment of the present invention, below with reference to Fig. 2 to above-mentioned steps S110-
S130 is described in detail.
Firstly, executing step S110, that is, obtain the face feature point of facial image.
The present embodiment can acquire video data by camera, using the image of collected every frame data as to be identified
Facial image.When acquiring image data, image data can be cut out, the image procossings such as noise reduction, it is subsequent to improve
To the recognition of face speed and precision of image data.
In some embodiments, the face feature point of facial image is obtained by following methods, comprising: firstly, using more
Task cascades convolutional network and carries out Face datection to the facial image, alternatively, using Machine learning tools Dlib to the people
Face image carries out Face datection, obtains the human face region subgraph of the facial image;Then, by the human face region subgraph
Trained convolutional neural networks are input to, the face feature point of the facial image is obtained.
The convolution of the foundation structures such as including MobileFaceNets, ShuffleNet, ResNet or VGG can be constructed in advance
Neural network, the present embodiment use Triplet, Arcface equal error calculation method training convolutional neural networks stage by stage, with
Recognition of face is carried out to trained convolutional neural networks.
During training convolutional neural networks, increase Asia face in data set, and clean, merging data collection;It is right
Facial image in data set carries out face feature point detection, carries out face correction, and utility data set can train to obtain
Good model;And the image concentrated to data is normalized, such as facial image size normalization to 112
× 112, pixel value subtract 127.5 after divided by 128.
After using data set training convolutional neural networks, the positive and negative test sample collection of certain scale, test volume are also set up
The classification performance of product neural network.The test set of foundation includes the library VIP, the facial image of n different users of storage, n in the library VIP
For the natural number greater than 50;The first test set is established, n facial images are left in the first test concentratedly, and this n opens facial image
User is the user registered in the library VIP;The second test set is established, n facial images are left in the second test concentratedly, this n opens
The user of facial image is the user registered not in the library VIP.
The image in image and the second test set in the first test set is identified with the convolutional neural networks that training obtains, and
People does not make face alignment with the library VIP, calculates human face similarity degree.Pass through each image and the library VIP face in two test sets of analysis
The distribution situation of human face similarity degree between image can detecte the classification performance of convolutional neural networks.
After the face feature point for obtaining facial image, step S120 is continued to execute, i.e., according to the face feature point
Facial angle, face width are obtained, and according to source figures multiple in the feature vector and image library being made of the facial features
The eigenmatrix of picture obtains the human face similarity degree between the facial image and the multiple source images.
In some embodiments, facial angle, face width are obtained by following methods: first according to the acquisition face
The camera internal parameter of image obtains match point of the face feature point in camera coordinates system, and the match point is described
Plane where in camera coordinates system is benchmark plane;Then 3D standard faces model, institute are constructed in the camera coordinates system
Stating includes multiple three-dimensional facial features points in 3D standard faces model;Then in the camera coordinates system, according to rotation peace
Multiple mapping points that the multiple three-dimensional facial features point is formed on the datum plane after shifting processing and the match point
It is overlapped the most rotation angle of quantity, obtains the facial angle;Finally according to both sides of the edge face in the face feature point
The distance between characteristic point obtains the face width.
It wherein, include: to the multiple three-dimensional facial features to the rotation and translation processing of multiple three-dimensional facial features points
Point rotated relative to XOY plane, to the multiple three-dimensional facial features point along the direction parallel with Y-axis carry out rotate and it is right
The multiple three-dimensional facial features point is rotated along the direction parallel with X-axis;Wherein, OXYZ is the camera coordinates system, O
For the origin of the camera coordinates system, X, Y and Z are respectively three reference axis of camera coordinates system, and the XOY plane is the base
Directrix plane.
The facial angle obtained as a result, includes: swing angle, rotational angle and pitch angle;Swing angle corresponds to more
A three-dimensional facial features point relative to XOY plane rotation angle, rotational angle correspond to multiple three-dimensional facial features points along with Y
The rotation angle of axis parallel direction, pitch angle correspond to multiple three-dimensional facial features points along the rotation with X-axis parallel direction
Angle.Swing angle, rotational angle and the pitch angle obtained at this time respectively corresponds course angle Yaw in Fig. 3, roll angle
Roll and pitch angle Pitch.
In practical applications, face feature point is subjected to coordinate mapping, is mapped to XOY plane in camera coordinates system, then
Multiple three-dimensional facial features points of 3D standard faces model in camera coordinates system are carried out to whole rotation and translation transformation, are made more
The mapping point of XOY plane is as much as possible in camera coordinates system and face feature point is in XOY plane for a three-dimensional facial features point
Mapping point be overlapped, the rotation angle carried out thus according to multiple face feature points can be obtained the facial angle of user.
In some embodiments, the face in facial image and image library between multiple source images is obtained by following methods
Similarity: it calculates separately in the feature vector and the eigenmatrix of the facial image between the feature vector of each source images
Cosine similarity, the maximum value in the cosine similarity calculated is determined as the facial image and the multiple source
Human face similarity degree between image.Wherein, the acquisition of the face feature point of the feature vector of each source images in image library is constituted
Method is to carry out Face datection to each source images using multitask concatenated convolutional network, alternatively, utilizing Machine learning tools
Dlib carries out Face datection to each source images, obtains the human face region subgraph of each source images;Then, by human face region
Image is input to trained convolutional neural networks, obtains the face feature point of each source images.
After obtaining facial angle, face width and human face similarity degree, step S130 is continued to execute, i.e., according to the people
Face angle, the face width and the human face similarity degree carry out recognition of face, obtain the recognition result of the facial image.
In some embodiments, the recognition result of facial image is obtained by following methods: by the facial angle and in advance
If angle threshold is compared, when the facial angle is less than the predetermined angle threshold value, identify that the facial image is nothing
Imitate image, when the facial angle is not less than the predetermined angle threshold value, by the face width and predetermined width threshold value into
Row compares, and when the face width is less than the predetermined width threshold value, identifies that the facial image is invalid image, at this time may be used
To generate the lesser prompt of face in facial image, when the face width is not less than the predetermined width threshold value, institute is obtained
The relationship of human face similarity degree and the first similarity threshold and the second similarity threshold is stated, first similarity threshold is greater than described
Second similarity threshold;According to the pass of the human face similarity degree and first similarity threshold and second similarity threshold
System, obtains the recognition result of facial image.
Wherein, if the human face similarity degree is greater than first similarity threshold, the corresponding use of the facial image is identified
Family be described image library in target source images owning user, the target source images be the multiple source images in the face
The corresponding source images of human face similarity degree maximum value between image;If the human face similarity degree is less than the second similarity threshold
Value identifies that the corresponding user of the facial image is new user, obtain the information of the new user and by the facial image and
The information of the new user is added in described image library;If the human face similarity degree no more than first similarity threshold and
The human face similarity degree is not less than second similarity threshold, reacquires the face feature point of facial image.
As shown in Fig. 2, in some embodiments, being identified based on Face datection result to facial image, detecting
When human face region is not present in facial image, recognition result is invalid image, as shown in Fig. 2, can export " No People " at this time
Prompt information;
Facial image is being detected there are when human face region, face is calculated according to the face feature point extracted from human face region
Angle is unsatisfactory for predetermined angle threshold value in facial angle, such as swing angle, rotational angle and pitch angle are respectively less than preset angle
When spending threshold value yz0, recognition result is invalid image;
When swing angle, rotational angle and pitch angle unspecified angle are greater than predetermined angle threshold value yz0, according to from face
The face feature point of extracted region calculates face width, and when face width is less than predetermined width threshold value yz1, recognition result is nothing
Image is imitated, as shown in Fig. 2, the prompt information of " Face too small " can be exported at this time.
When face width is not less than predetermined width threshold value yz1, further compare the people between facial image and source images
Face similarity degree, when human face similarity degree is greater than the first similarity threshold yz2, the corresponding user of identification facial image is described image
Target source images owning user in library, as shown in Fig. 2, the prompt information of user name and human face similarity degree can be exported at this time;
When human face similarity degree is less than the second similarity threshold yz3, the corresponding user of identification facial image is new user, is obtained
It takes the information of the new user and the information of the facial image and the new user is added in described image library, such as Fig. 2
It is shown, the prompt information of " New People " can be exported at this time;
It is not less than the second similarity threshold yz3 in human face similarity degree and human face similarity degree is not more than the first similarity threshold
When yz2, recognition result is recognition failures, as shown in Fig. 2, can export the prompt information of " Try again " at this time.
The face identification method speed advantage of the present embodiment is obvious, can run in a mobile device, such as mobile phone
Etc. mobile devices.
The present invention also provides a kind of cameras.
Fig. 4 is the structural block diagram of the camera shown in the embodiment of the present invention, as shown in figure 4, the camera of the present embodiment includes: to take the photograph
As head and processor;
The camera acquires facial image and is sent to processor;
The processor obtains the face feature point of facial image;Facial angle, people are obtained according to the face feature point
Face width, and according to the eigenmatrix of multiple source images in the feature vector and image library being made of the facial features, obtain
Take the human face similarity degree between the facial image and the multiple source images;According to the facial angle, the face width
Recognition of face is carried out with the human face similarity degree, obtains the recognition result of the facial image.
The present embodiment can acquire video data by camera, using the image of collected every frame data as to be identified
Facial image.When acquiring image data, image data can be cut out, the image procossings such as noise reduction, it is subsequent to improve
To the recognition of face speed and precision of image data.
In some embodiments, the facial angle is compared, in the face by processor with predetermined angle threshold value
It when angle is less than the predetermined angle threshold value, identifies that the facial image is invalid image, is not less than institute in the facial angle
When stating predetermined angle threshold value, the face width is compared with predetermined width threshold value, is less than in the face width described
It when predetermined width threshold value, identifies that the facial image is invalid image, is not less than the predetermined width threshold in the face width
When value, the relationship of the human face similarity degree and the first similarity threshold and the second similarity threshold, first similarity are obtained
Threshold value is greater than second similarity threshold;If the human face similarity degree is greater than first similarity threshold, the people is identified
The corresponding user of face image is target source images owning user in described image library, and the target source images are the multiple source figure
The corresponding source images of human face similarity degree maximum value between the facial image as in;If the human face similarity degree is less than described
Second similarity threshold identifies that the corresponding user of the facial image is new user, obtains the information of the new user and by institute
The information for stating facial image and the new user is added in described image library;If the human face similarity degree is not more than described first
Similarity threshold and the human face similarity degree are not less than second similarity threshold, reacquire the facial characteristics of facial image
Point.
In some embodiments, processor obtains the face also according to the camera internal parameter for acquiring the facial image
Match point of portion's characteristic point in camera coordinates system, the match point are flat on the basis of the plane at place in the camera coordinates system
Face;3D standard faces model is constructed in the camera coordinates system, includes multiple three dimensional faces in the 3D standard faces model
Characteristic point;In the camera coordinates system, the multiple three-dimensional facial features point is in the base after being handled according to rotation and translation
The multiple mapping points formed on directrix plane are overlapped the most rotation angle of quantity with the match point, obtain the face angle
Degree;According to the distance between both sides of the edge face feature point in the face feature point, the face width is obtained.
In some embodiments, processor carries out face inspection to the facial image using multitask concatenated convolutional network
It surveys, alternatively, carrying out Face datection to the facial image using Machine learning tools Dlib, obtains the face of the facial image
Region subgraph;The human face region subgraph is input to trained convolutional neural networks, obtains the facial image
Face feature point.
For camera embodiment, since it corresponds essentially to embodiment of the method, so related place is referring to method reality
Apply the part explanation of example.The apparatus embodiments described above are merely exemplary, wherein described be used as separation unit
The unit of explanation may or may not be physically separated, and component shown as a unit can be or can also be with
It is not physical unit, it can it is in one place, or may be distributed over multiple network units.It can be according to actual
It needs that some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not
In the case where making the creative labor, it can understand and implement.
The above description is merely a specific embodiment, under above-mentioned introduction of the invention, those skilled in the art
Other improvement or deformation can be carried out on the basis of the above embodiments.It will be understood by those skilled in the art that above-mentioned tool
Body description only preferably explains that the purpose of the present invention, protection scope of the present invention should be subject to the protection scope in claims.
Claims (10)
1. a kind of face identification method characterized by comprising
Obtain the face feature point of facial image;
Facial angle, face width are obtained according to the face feature point, and according to the feature being made of the facial features
The eigenmatrix of multiple source images, obtains the face between the facial image and the multiple source images in vector and image library
Similarity;
Recognition of face is carried out according to the facial angle, the face width and the human face similarity degree, obtains the face figure
The recognition result of picture.
2. the method according to claim 1, wherein it is described according to the facial angle, the face width and
The human face similarity degree carries out recognition of face, obtains the recognition result of the facial image, comprising:
The facial angle is compared with predetermined angle threshold value, is less than the predetermined angle threshold value in the facial angle
When, identify that the facial image is invalid image, when the facial angle is not less than the predetermined angle threshold value,
The face width is compared with predetermined width threshold value, is less than the predetermined width threshold value in the face width
When, identify that the facial image is invalid image, when the face width is not less than the predetermined width threshold value, described in acquisition
The relationship of human face similarity degree and the first similarity threshold and the second similarity threshold, first similarity threshold are greater than described the
Two similarity thresholds;
According to the relationship of the human face similarity degree and first similarity threshold and second similarity threshold, face is obtained
The recognition result of image.
3. according to the method described in claim 2, it is characterized in that, described similar to described first according to the human face similarity degree
The relationship for spending threshold value and second similarity threshold, obtains the recognition result of facial image, comprising:
If the human face similarity degree is greater than first similarity threshold, identify that the corresponding user of the facial image is the figure
As target source images owning user in library, the target source images be in the multiple source images between the facial image
The corresponding source images of human face similarity degree maximum value;
If the human face similarity degree is less than second similarity threshold, identify that the corresponding user of the facial image is new uses
Family obtains the information of the new user and the information of the facial image and the new user is added in described image library;
If the human face similarity degree is no more than first similarity threshold and the human face similarity degree is not less than second phase
Like degree threshold value, the face feature point of facial image is reacquired.
4. the method according to claim 1, wherein the face feature point for obtaining facial image, comprising:
Face datection is carried out to the facial image using multitask concatenated convolutional network, alternatively, utilizing Machine learning tools
Dlib carries out Face datection to the facial image, obtains the human face region subgraph of the facial image;
The human face region subgraph is input to trained convolutional neural networks, obtains the facial characteristics of the facial image
Point.
5. the method according to claim 1, wherein it is described according to the face feature point obtain facial angle,
Face width, comprising:
According to the camera internal parameter for acquiring the facial image, matching of the face feature point in camera coordinates system is obtained
Point, plane of the match point where in the camera coordinates system are benchmark plane;
3D standard faces model is constructed in the camera coordinates system, includes multiple three dimensional faces in the 3D standard faces model
Characteristic point;
In the camera coordinates system, the multiple three-dimensional facial features point is flat in the benchmark after being handled according to rotation and translation
The multiple mapping points formed on face are overlapped the most rotation angle of quantity with the match point, obtain the facial angle;
According to the distance between both sides of the edge face feature point in the face feature point, the face width is obtained.
6. according to the method described in claim 5, it is characterized in that, rotation and translation processing includes:
The multiple three-dimensional facial features point is rotated relative to XOY plane, to the multiple three-dimensional facial features point along
The direction parallel with Y-axis rotate and rotate to the multiple three-dimensional facial features point along the direction parallel with X-axis;
Wherein, OXYZ is the camera coordinates system, and O is the origin of the camera coordinates system, and X, Y and Z are respectively camera coordinates system
Three reference axis, the XOY plane be the datum plane.
7. according to the method described in claim 6, it is characterized in that, the facial angle include: swing angle, rotational angle and
Pitch angle;The swing angle correspond to the multiple three-dimensional facial features point relative to XOY plane rotation angle, described turn
Dynamic angle corresponds to the multiple three-dimensional facial features point along the rotation angle with Y-axis parallel direction, the pitch angle pair
It should be in the multiple three-dimensional facial features point along the rotation angle with X-axis parallel direction.
8. the method according to claim 1, wherein the feature vector that the basis is made of the facial features
With the eigenmatrix of source images multiple in image library, the face obtained between the facial image and the multiple source images is similar
Degree, comprising:
It calculates separately in the feature vector and the eigenmatrix of the facial image between the feature vector of each source images
Maximum value in the cosine similarity calculated is determined as the facial image and the multiple source figure by cosine similarity
Human face similarity degree as between.
9. a kind of camera characterized by comprising camera and processor;
The camera acquires facial image and is sent to the processor;
The processor obtains the face feature point of facial image;It is wide that facial angle, face are obtained according to the face feature point
Degree, and according to the eigenmatrix of multiple source images in the feature vector and image library being made of the facial features, obtain institute
State the human face similarity degree between facial image and the multiple source images;According to the facial angle, the face width and institute
It states human face similarity degree and carries out recognition of face, obtain the recognition result of the facial image.
10. camera according to claim 9, which is characterized in that
The facial angle is compared by the processor with predetermined angle threshold value, is less than in the facial angle described pre-
If when angle threshold, identifying that the facial image is invalid image, it is not less than the predetermined angle threshold value in the facial angle
When, the face width is compared with predetermined width threshold value, when the face width is less than the predetermined width threshold value,
It identifies that the facial image is invalid image, when the face width is not less than the predetermined width threshold value, obtains the people
The relationship of face similarity degree and the first similarity threshold and the second similarity threshold, first similarity threshold are greater than described second
Similarity threshold;If the human face similarity degree is greater than first similarity threshold, the corresponding user of the facial image is identified
For target source images owning user in described image library, the target source images be in the multiple source images with the face figure
The corresponding source images of human face similarity degree maximum value as between;If the human face similarity degree is less than second similarity threshold,
It identifies that the corresponding user of the facial image is new user, obtains the information of the new user and by the facial image and described
The information of new user is added in described image library;If the human face similarity degree is no more than first similarity threshold and described
Human face similarity degree is not less than second similarity threshold, reacquires the face feature point of facial image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910395391.4A CN110188630A (en) | 2019-05-13 | 2019-05-13 | A kind of face identification method and camera |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910395391.4A CN110188630A (en) | 2019-05-13 | 2019-05-13 | A kind of face identification method and camera |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110188630A true CN110188630A (en) | 2019-08-30 |
Family
ID=67716090
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910395391.4A Pending CN110188630A (en) | 2019-05-13 | 2019-05-13 | A kind of face identification method and camera |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110188630A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110837773A (en) * | 2019-09-27 | 2020-02-25 | 深圳市华付信息技术有限公司 | Large-angle face pose estimation method based on deep learning |
CN110843726A (en) * | 2019-12-19 | 2020-02-28 | 北京明朝万达科技股份有限公司 | Automobile door lock control system and method and vehicle |
CN113536844A (en) * | 2020-04-16 | 2021-10-22 | 中移(成都)信息通信科技有限公司 | Face comparison method, device, equipment and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107066969A (en) * | 2017-04-12 | 2017-08-18 | 南京维睛视空信息科技有限公司 | A kind of face identification method |
CN107145842A (en) * | 2017-04-19 | 2017-09-08 | 西安电子科技大学 | With reference to LBP characteristic patterns and the face identification method of convolutional neural networks |
CN107194341A (en) * | 2017-05-16 | 2017-09-22 | 西安电子科技大学 | The many convolution neural network fusion face identification methods of Maxout and system |
CN108549873A (en) * | 2018-04-19 | 2018-09-18 | 北京华捷艾米科技有限公司 | Three-dimensional face identification method and three-dimensional face recognition system |
-
2019
- 2019-05-13 CN CN201910395391.4A patent/CN110188630A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107066969A (en) * | 2017-04-12 | 2017-08-18 | 南京维睛视空信息科技有限公司 | A kind of face identification method |
CN107145842A (en) * | 2017-04-19 | 2017-09-08 | 西安电子科技大学 | With reference to LBP characteristic patterns and the face identification method of convolutional neural networks |
CN107194341A (en) * | 2017-05-16 | 2017-09-22 | 西安电子科技大学 | The many convolution neural network fusion face identification methods of Maxout and system |
CN108549873A (en) * | 2018-04-19 | 2018-09-18 | 北京华捷艾米科技有限公司 | Three-dimensional face identification method and three-dimensional face recognition system |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110837773A (en) * | 2019-09-27 | 2020-02-25 | 深圳市华付信息技术有限公司 | Large-angle face pose estimation method based on deep learning |
CN110843726A (en) * | 2019-12-19 | 2020-02-28 | 北京明朝万达科技股份有限公司 | Automobile door lock control system and method and vehicle |
CN113536844A (en) * | 2020-04-16 | 2021-10-22 | 中移(成都)信息通信科技有限公司 | Face comparison method, device, equipment and medium |
CN113536844B (en) * | 2020-04-16 | 2023-10-31 | 中移(成都)信息通信科技有限公司 | Face comparison method, device, equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11749025B2 (en) | Eye pose identification using eye features | |
CN111243093B (en) | Three-dimensional face grid generation method, device, equipment and storage medium | |
WO2020228389A1 (en) | Method and apparatus for creating facial model, electronic device, and computer-readable storage medium | |
WO2022089360A1 (en) | Face detection neural network and training method, face detection method, and storage medium | |
CN109410316B (en) | Method for three-dimensional reconstruction of object, tracking method, related device and storage medium | |
CN113689503B (en) | Target object posture detection method, device, equipment and storage medium | |
CN110188630A (en) | A kind of face identification method and camera | |
CN109766866B (en) | Face characteristic point real-time detection method and detection system based on three-dimensional reconstruction | |
CN109858433B (en) | Method and device for identifying two-dimensional face picture based on three-dimensional face model | |
WO2024012333A1 (en) | Pose estimation method and apparatus, related model training method and apparatus, electronic device, computer readable medium and computer program product | |
US10791321B2 (en) | Constructing a user's face model using particle filters | |
CN110046554B (en) | Face alignment method and camera | |
US9786030B1 (en) | Providing focal length adjustments | |
CN115344113A (en) | Multi-view human motion capture method, device, system, medium and terminal | |
CN118097027A (en) | Multi-video 3D human motion capturing method and system based on evolutionary computation | |
WO2020181509A1 (en) | Image processing method, apparatus and system | |
Wang et al. | Design of Real Time Display Algorithm for Naked Eye 3D Images Based on Binocular Vision | |
CN115880428A (en) | Animal detection data processing method, device and equipment based on three-dimensional technology | |
CN116883472A (en) | Face nursing system based on face three-dimensional image registration | |
CN116994308A (en) | Method and device for detecting lip synchronization result, electronic equipment and storage medium | |
Li et al. | Panoramic face and ear image stitching in multi-modal recognition | |
CN110188604A (en) | Face identification method and device based on 2D and 3D image |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190830 |