CN109858433A - A kind of method and device based on three-dimensional face model identification two-dimension human face picture - Google Patents
A kind of method and device based on three-dimensional face model identification two-dimension human face picture Download PDFInfo
- Publication number
- CN109858433A CN109858433A CN201910082406.1A CN201910082406A CN109858433A CN 109858433 A CN109858433 A CN 109858433A CN 201910082406 A CN201910082406 A CN 201910082406A CN 109858433 A CN109858433 A CN 109858433A
- Authority
- CN
- China
- Prior art keywords
- face
- image
- dimensional
- feature vector
- dimension human
- 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.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of face identification methods based on three-dimensional face model identification two-dimension human face picture, comprising: is pre-processed to obtain characteristic point, the attitude angle of facial image to be identified to facial image to be identified；Alignment, outer registration process in plane are carried out, respectively to facial image to be identified to obtain the first Face Image with Pose Variations collection；The feature vector of the first Face Image with Pose Variations collection is extracted by convolutional neural networks, and seeks its averaged feature vector；Three-dimensional face data collection is established, inside and outside registration process is carried out to each three-dimensional face model that three-dimensional face data is concentrated, generates the second Face Image with Pose Variations collection of each model；The feature vector of the second Face Image with Pose Variations collection is extracted by convolutional neural networks, and finds out the second averaged feature vector；The first averaged feature vector and the second averaged feature vector are compared, face recognition result is obtained.This method carries out two-dimension human face image identification using the abundant posture information that three-dimensional face contains, and further increases face recognition accuracy.
Description
Technical field
The present invention relates to computer visions and mode identification technology, more particularly to a kind of three-dimensional face model that is based on to know
The method and device of other two-dimension human face picture.
Background technique
Face recognition technology has been daily life band as one of the biometrics identification technology being widely spread
It is huge convenient to come.Traditional face recognition technology is the recognition of face based on two-dimentional (2D) picture, that is, registers end and identification end all
Face 2D image is acquired, face characteristic is extracted later and is compared, complete the identification or data of face.Current 2D recognition of face
(illumination good, posture face) achieves preferable recognition result in limited conditions, and recognition effect under the conditions of non-limiting
Will sharply it decline, such as big attitudes vibration is presented in face.2D face recognition technology has gradually tended to be mature and has reached at present
Certain bottleneck introduces more characteristic informations to the development need of face recognition technology to overcome the shortcomings of 2D recognition of face.Base
In one of the trend that the face recognition technology of three-dimensional (3D) faceform is future development, 3D people's model ratio 2D face picture possesses
The information more abundant such as 3D shape, can be improved recognition performance.However all 3D is used to pass at registration end and identification end
Sensor captures 3D face, and 2D camera all on present society is all transformed into 3D sensor being then cannot be real in the short time
It is existing.A kind of practical plan is registration end acquisition 3D face, and identifies that end acquisition 2D face is identified, relates to the use of three-dimensional face
Model identifies the relevant technologies of two-dimension human face picture, this is also the background that the invention patent proposes.
The technology for relating to the use of three-dimensional face model identification two-dimension human face image at present is more deficient, and majority also rests on benefit
The stage of three-dimensional face is identified with three-dimensional face model, and non-used three-dimensional face model removes identification two-dimension human face image.Application
The Chinese invention patent application that publication No. is CN108427871A discloses a kind of 3D face rapid identity authentication method and device,
Three-dimensional face model is rotated to the identical posture of two dimensional image to be identified and projects to two dimensional image by it, then by two after projection
It ties up image and identification is compared with two dimensional image to be identified.
Threedimensional model is only projected into the corresponding single posture of two dimensional image however, existing, without utilizing three-dimensional face
The abundant posture information contained carries out the technical issues of two-dimension human face image identification.
Summary of the invention
An object of the present invention at least that, for how to overcome the above-mentioned problems of the prior art, provide one kind
Based on the method and device of three-dimensional face model identification two-dimension human face picture, the abundant posture that can be provided using three-dimensional face is believed
Breath carries out two-dimension human face image identification, improves and only uses single gesture recognition in can not identifying of being likely to occur of complex environment or know
Not wrong problem, improves the robustness and fault-tolerant ability of system, and due to the bright-dark degree of the method for use and photo without
It closes, can solve in face recognition process is influenced by shooting environmental intensity of illumination, therefore be can achieve comparatively ideal face and known
Other effect, effectively increases face recognition accuracy rate, so that recognition of face has more practicability.
To achieve the goals above, the technical solution adopted by the present invention includes following aspects.
A kind of face identification method based on three-dimensional face model identification two-dimension human face picture, comprising:
Step 101, a certain number of facial images to be identified are obtained, the facial image to be identified pre-process
To the characteristic point of facial image to be identified, and obtain the attitude angle of the facial image to be identified；
Step 102, registration process outside registration process and plane is carried out respectively in plane to the facial image to be identified, obtained
The first Face Image with Pose Variations collection after to alignment；The first Face Image with Pose Variations collection is facial image to be identified by flat
The two-dimension human face image collection that registration process and the outer registration process of plane obtain in face comprising multiple faces are under multiple postures
Two-dimension human face image；
Step 103, the feature vector of the first Face Image with Pose Variations collection is extracted by convolutional neural networks, and is acquired
First averaged feature vector of image of each of first Face Image with Pose Variations collection face under multiple postures；
Step 104, three-dimensional face data collection is obtained, each of three-dimensional face data collection three-dimensional face model is passed through
It is aligned outside plane and is aligned to different postures in plane, generate the second Face Image with Pose Variations collection comprising each is three-dimensional
Two dimensional image of the faceform under multiple postures；
Step 105, the feature vector of the second Face Image with Pose Variations collection is extracted by convolutional neural networks, and is found out
Second average characteristics of two dimensional image of each three-dimensional face model that the second Face Image with Pose Variations is concentrated under multiple postures
Vector；
Step 106, by the first averaged feature vector obtained according to facial image to be identified and according to three-dimensional face data
The second averaged feature vector for concentrating each three-dimensional face model to obtain compares, and obtains face recognition result.
Preferably, in a kind of face identification method based on three-dimensional face model identification two-dimension human face picture, the posture
Angle is yaw angle.
Preferably, described to institute in a kind of face identification method based on three-dimensional face model identification two-dimension human face picture
The two-dimension human face area image for stating facial image to be identified carries out registration process in plane and includes:
Determine the two-dimension human face area image characteristic point coordinate to template point coordinate similarity transformation relationship, and obtain into
Two-dimension human face image after row similarity transformation.
Preferably, described to institute in a kind of face identification method based on three-dimensional face model identification two-dimension human face picture
The two-dimension human face area image for stating facial image to be identified carries out the outer registration process of plane and includes:
The two-dimension human face area image is generated into threedimensional model, projection function is determined according to the attitude angle, is based on institute
It states projection function and the threedimensional model of generation is projected into corresponding two-dimension human face image according to posture.
Preferably, in a kind of face identification method based on three-dimensional face model identification two-dimension human face picture, the step
104 specifically include:
Obtain three-dimensional face data collection, by the three-dimensional face model that the three-dimensional face data is concentrated rotate to it is corresponding to
The attitude angle for identifying facial image, projects to two dimensional image for postrotational three-dimensional face model, and according to corresponding characteristic point
Registration process in plane is done to resulting two dimensional image is projected；And registration process outside plane is done to the three-dimensional face model, i.e.,
Three-dimensional face model is projected into corresponding two-dimension human face image according to the attitude angle, to generate the second multi-pose after alignment
Face image set.
Preferably, in a kind of face identification method based on three-dimensional face model identification two-dimension human face picture, the convolution
Neural network is one of Inception-v4, Inception-Resnet-v1, Inception-Resnet-v2.
Preferably, in a kind of face identification method based on three-dimensional face model identification two-dimension human face picture, pass through calculating
Cosine similarity or Euclidean distance between vector carry out the comparison of the first averaged feature vector and the second averaged feature vector.
A kind of device based on three-dimensional face model identification two-dimension human face picture, including at least one processor, Yi Jiyu
The memory of at least one processor communication connection；The memory, which is stored with, to be executed by least one described processor
Instruction, described instruction is executed by least one described processor, so that at least one described processor is able to carry out above-mentioned side
Method.
In conclusion by adopting the above-described technical solution, the present invention at least has the advantages that
Two-dimension human face image identification is carried out by the abundant posture information contained using three-dimensional face, three-dimensional people can be utilized
The abundant posture information that face provides carries out two-dimension human face image identification, and it is possible in complex environment that improvement only uses single gesture recognition
What is occurred can not identify or identify the problem of mistake, improve the robustness and fault-tolerant ability of system, and due to the method for use
Unrelated with the bright-dark degree of photo, can solve in face recognition process is influenced by shooting environmental intensity of illumination, therefore can be with
Reach comparatively ideal recognition of face effect, effectively increase face recognition accuracy rate, so that recognition of face has more practicability.
Detailed description of the invention
Fig. 1 is that the face according to an exemplary embodiment of the present invention based on three-dimensional face model identification two-dimension human face picture is known
Other method flow diagram.
Fig. 2 is that the face according to an exemplary embodiment of the present invention based on three-dimensional face model identification two-dimension human face picture is known
Other method schematic.
Fig. 3 is three-dimensional face model human face characteristic point schematic diagram according to an exemplary embodiment of the present invention.
Fig. 4 is the device knot according to an exemplary embodiment of the present invention based on three-dimensional face model identification two-dimension human face picture
Structure schematic diagram.
Specific embodiment
With reference to the accompanying drawings and embodiments, the present invention will be described in further detail, so that the purpose of the present invention, technology
Scheme and advantage are more clearly understood.It should be appreciated that described herein, specific examples are only used to explain the present invention, and does not have to
It is of the invention in limiting.
Fig. 1, Fig. 2 shows according to an exemplary embodiment of the present invention a kind of based on three-dimensional face model identification two-dimension human face
The face identification method of picture.The method of the embodiment specifically includes that
Step 101, a certain number of facial images to be identified are obtained, the facial image to be identified pre-process
To the human face characteristic point of facial image to be identified, and obtain the attitude angle of the facial image to be identified；
Specifically, facial image to be identified can be the two-dimension human face picture of screening acquisition (in three-dimensional face database
Three-dimensional face model it is corresponding), be also possible to the two-dimension picture by candid photographs such as monitoring, cameras.Therefore, images to be recognized is
Two dimensional image, and be three-dimensional face model inside image library, the recognition of face problem in the present embodiment is to answer images to be recognized
Belong to the identity of which three-dimensional face model inside library, i.e., identifies two-dimension human face image with three-dimensional face model.
The facial image to be identified is pre-processed, wherein pretreatment includes: to utilize Face datection algorithm detection two
Tieing up facial image region, (method for extracting two-dimension human face image region is relatively more, we use document " Joint Face here
Detection and Alignment using Multi-task Cascaded Convolutional Networks " in
The detection in Face datection algorithm progress two-dimension human face image region)；It is extracted in images to be recognized with facial feature points detection algorithm
Two-dimension human face region human face characteristic point (extract two-dimension human face image provincial characteristics point method it is relatively more, we adopt here
With document " How far are we from solving the 2D&3D Face Alignment problem? (and a
230,000 3D facial landmarks of dataset of) " propose facial feature points detection algorithm obtain face 68
A characteristic point).
And estimate the attitude angle in two-dimension human face image region in the facial image to be identified(here using text
Offer the appearance of the method estimation face of " Fine-Grained Head Pose Estimation Without Keypoints " proposition
State angle), wherein ω indicates yaw angle (yaw), and θ indicates roll angle (roll),It indicates pitch angle (pitch).Simultaneously because people
Face identification is generally influenced influence the most obvious, therefore that ω is only considered in the present embodiment by yaw angle (yaw) ω, i.e., the described posture
Angle is yaw angle.And due to the symmetry of recognition of face, only the case where consideration ω >=0 °, if image yaw angle is discontented at this time
Sufficient ω >=0 ° can carry out flip horizontal to the two-dimension human face area image then to make its yaw angle meet ω >=0 °.
Step 102, registration process outside registration process and plane is carried out respectively in plane to the facial image to be identified, obtained
The first Face Image with Pose Variations collection after to alignment；First Face Image with Pose Variations integrates as registration process in plane and plane
The two-dimension human face image collection that outer registration process obtains comprising two-dimension human face image of multiple faces under multiple postures；
Specifically, to registration process outside registration process in every facial image progress plane to be identified and plane, and put down
Alignment and the outer registration process of plane can carry out simultaneously in face, can also operate in certain sequence, not influence operating result.It is described
Registration process is to calculate the similarity transformation of characteristic point (Fig. 3 is human face characteristic point schematic diagram) coordinate to template point coordinate in plane,
Similarity transformation is carried out to human face region using the transformation relation.Input human face region image is indicated with I, with functionIt indicates
According to attitude angleDetermining similarity transformation operation, transformed image are expressed as
In above formula, when the yaw angle ω of face is less than 45 degree, it is aligned using nine relatively stable characteristic points of front；
When the yaw angle of face is bigger, then using 3 characteristic points --- i.e. two centers and this 3 characteristic points calculating of nose are similar
Transformation.Features above point can be directly acquired from 68 characteristic points of face.In addition, corresponding with template point according to characteristic point
It is method well known to those skilled in the art that relationship, which calculates corresponding similarity transformation,.
Further, described that registration process outside plane is carried out to facial image to be identified, that is, it is based on generating three from two dimensional image
The method of dimension module, and threedimensional model generated is projected into two dimensional image according to posture.Specifically, it is indicated with F by single width
The method that two dimensional image generates threedimensional model indicates that input human face region image, the three-dimensional face model of generation are expressed as F with I
(I)；With functionIt indicates according to attitude angleDetermining projection function, the result table after threedimensional model projection
It is shown asIn the present embodiment, F uses document " A Morphable Model For The Synthesis Of
3D Faces " propose 3DMM method from single width two dimensional image generate threedimensional model, the projecting method of useDesign is such as
Under:
Thus the first Face Image with Pose Variations collection after being aligned；First Face Image with Pose Variations integrates as in plane
The set for the two-dimension human face image that registration process and the outer registration process of plane obtain comprising multiple faces are under multiple postures
Two-dimension human face image.
Step 103, the feature vector of the first Face Image with Pose Variations collection is extracted by convolutional neural networks, and is acquired
First averaged feature vector of image of each of first Face Image with Pose Variations collection face under multiple postures；
Specifically, by a series of multi-poses projected image obtained in step 102 (the first Face Image with Pose Variations collection) point
Face feature vector is not extracted by corresponding convolutional neural networks, feature vector obtained by each face is averagely obtained final
Feature vector.The convolutional neural networks are that common depth characteristic extracts neural network Inception-v4, Inception-
One of Resnet-v1, Inception-Resnet-v2, the depth characteristic are extracted neural network and are used for people's image face
Data carry out multiple residual error and add process of convolution to extract the depth characteristic vector of image.And the corresponding convolutional Neural of a certain posture
Network should be by the optimization training of sample under the posture, so that corresponding convolutional neural networks can extract image under the posture
Depth characteristic.
Step 104, three-dimensional face data collection is obtained, each of three-dimensional face data collection three-dimensional face model is passed through
It is aligned outside plane and is aligned to different postures in plane, generate the second Face Image with Pose Variations collection comprising each is three-dimensional
Two dimensional image of the faceform under multiple postures；
Specifically, obtain a certain number of three-dimensional face models (collected by special three-dimensional image acquisition equipment, or
Obtained from corresponding registry), to establish corresponding three-dimensional face data collection.Each three-dimensional face mould that data are concentrated
Type outside plane by being aligned and being aligned to different postures, i.e., the three-dimensional face mould concentrated the three-dimensional face data in plane
Type is rotated to the attitude angle of corresponding facial image to be identified, and postrotational three-dimensional face model is projected to two dimensional image, and
Registration process in plane is done to resulting two dimensional image is projected according to corresponding characteristic point；And the three-dimensional face model is done flat
Three-dimensional face model is projected to corresponding two-dimension human face image according to the attitude angle by registration process outside face, last gained
By in plane alignment, outer registration process two-dimension human face image collection be aligned after the second Face Image with Pose Variations collection,
And the second Face Image with Pose Variations collection includes two dimensional image of each three-dimensional face model under multiple postures.
Wherein interior registration process includes: and first rotates threedimensional model G to the attitude angle of face to be identifiedIt throws
For shadow to two dimensional image, projected image is expressed as I', and is done in plane and be aligned according to characteristic point.To keep the consistent of recognition of face
Property, using alignment thereof in plane identical with step 102That is formula (1).
Outer registration process includes: by threedimensional model G using projection pattern identical with step 102I.e. formula (2) is thrown
For shadow to two dimensional image, which is the second Face Image with Pose Variations collection.Three-dimensional face data is concentrated each
A three-dimensional face model collectively forms the second Face Image with Pose Variations by interior processing, the outer resulting two-dimension human face image collection of processing
Collection.
Step 105, the feature vector of the second Face Image with Pose Variations collection is extracted by convolutional neural networks, and is found out
Second average characteristics of two dimensional image of each three-dimensional face model that the second Face Image with Pose Variations is concentrated under multiple postures
Vector；
Specifically, using the convolutional neural networks in step 103 to the second Face Image with Pose Variations obtained in step 104
Collection carries out feature extraction, obtains the second Face Image with Pose Variations and concentrates the corresponding two dimensional image of each three-dimensional face model (outer right
Together, interior to be aligned obtained multi-pose two dimensional image) averaged feature vector.
Step 106, by the first averaged feature vector obtained according to facial image to be identified and according to three-dimensional face data
The second averaged feature vector for concentrating each three-dimensional face model to obtain compares, and obtains face recognition result.
Specifically, by the first averaged feature vector handled face image set to be identified and according to three-dimensional people
The second averaged feature vector that face data set obtains compares, and obtains face recognition result.Obtain the first average vector and
After two average vectors, vector is compared, obtains recognition result.Face feature vector comparison is the step of field is often used,
It can be compared by calculating cosine similarity between vector or Euclidean distance.And when a certain in three-dimensional face data
The feature vector of a three-dimensional face model and the feature vector similarity maximum or the two of some facial image to be identified are European
Apart from it is nearest when, then the three-dimensional face model is the recognition result of corresponding two-dimension human face image to be identified.
In further embodiment of the present invention, step 102~step 103 (obtains two-dimension human face image data to be identified
Averaged feature vector the step of) with step 104~step 105 (obtain three-dimensional face data collection averaged feature vector step
Suddenly it successively can sequentially be operated, can also be extracted simultaneously), operation order is in no particular order.
In above-described embodiment, two-dimension human face image identification is carried out by the abundant posture information contained using three-dimensional face,
The abundant posture information that can be provided using three-dimensional face carries out two-dimension human face image identification, and improvement only uses single gesture recognition
In the problem that can not identify or identify mistake that complex environment is likely to occur, the robustness and fault-tolerant ability of system are improved, and
Since the method for use and the bright-dark degree of photo are unrelated, can solve in face recognition process by shooting environmental intensity of illumination
It influences, therefore can achieve comparatively ideal recognition of face effect, face recognition accuracy rate is effectively increased, so that recognition of face is more
With practicability.
Fig. 4 shows according to an exemplary embodiment of the present invention a kind of based on three-dimensional face model identification two-dimension human face picture
Device, i.e., electronic equipment 310 (such as have program execute function computer server) comprising at least one processor
311, power supply 314, and memory 312 and input/output interface 313 with the communication connection of at least one described processor 311；
The memory 312 is stored with the instruction that can be executed by least one described processor 311, described instruction by it is described at least one
Processor 311 executes, so that at least one described processor 311 is able to carry out method disclosed in aforementioned any embodiment；Institute
Stating input/output interface 313 may include display, keyboard, mouse and USB interface, be used for inputoutput data；Power supply
314 for providing electric energy for electronic equipment 310.
It will be appreciated by those skilled in the art that: realize that all or part of the steps of above method embodiment can pass through program
Relevant hardware is instructed to complete, program above-mentioned can store in computer-readable storage medium, which is executing
When, execute step including the steps of the foregoing method embodiments；And storage medium above-mentioned includes: movable storage device, read-only memory
The various media that can store program code such as (Read Only Memory, ROM), magnetic or disk.
When the above-mentioned integrated unit of the present invention be realized in the form of SFU software functional unit and as the sale of independent product or
In use, also can store in a computer readable storage medium.Based on this understanding, the skill of the embodiment of the present invention
Substantially the part that contributes to existing technology can be embodied in the form of software products art scheme in other words, the calculating
Machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be individual
Computer, server or network equipment etc.) execute all or part of each embodiment the method for the present invention.And it is aforementioned
Storage medium include: the various media that can store program code such as movable storage device, ROM, magnetic or disk.
The above, the only detailed description of the specific embodiment of the invention, rather than limitation of the present invention.The relevant technologies
The technical staff in field is not in the case where departing from principle and range of the invention, various replacements, modification and the improvement made
It should all be included in the protection scope of the present invention.
Claims (8)
1. a kind of face identification method based on three-dimensional face model identification two-dimension human face picture, which is characterized in that described to include:
Step 101, a certain number of facial images to be identified are obtained, to the facial image to be identified pre-processed to obtain to
It identifies the characteristic point of facial image, and obtains the attitude angle of the facial image to be identified；
Step 102, registration process outside registration process and plane is carried out respectively in plane to the facial image to be identified, obtained pair
The first Face Image with Pose Variations collection after neat；The first Face Image with Pose Variations collection is facial image to be identified by plane
The two-dimension human face image collection that registration process and the outer registration process of plane obtain comprising two dimension of multiple faces under multiple postures
Facial image；
Step 103, the feature vector of the first Face Image with Pose Variations collection is extracted by convolutional neural networks, and acquires first
First averaged feature vector of image of each of the Face Image with Pose Variations collection face under multiple postures；
Step 104, three-dimensional face data collection is obtained, plane is passed through to each of three-dimensional face data collection three-dimensional face model
It is aligned to different postures in outer alignment and plane, generates the second Face Image with Pose Variations collection comprising each three-dimensional face
Two dimensional image of the model under multiple postures；
Step 105, the feature vector of the second Face Image with Pose Variations collection is extracted by convolutional neural networks, and finds out second
Second averaged feature vector of two dimensional image of each three-dimensional face model that Face Image with Pose Variations is concentrated under multiple postures；
Step 106, it is concentrated by the first averaged feature vector obtained according to facial image to be identified and according to three-dimensional face data
The second averaged feature vector that each three-dimensional face model obtains compares, and obtains face recognition result.
2. the method according to claim 1, wherein the attitude angle is yaw angle.
3. the method according to claim 1, wherein the two-dimension human face area to the facial image to be identified
Area image carries out registration process in plane
Determine the two-dimension human face area image characteristic point coordinate to template point coordinate similarity transformation relationship, and obtain carry out phase
Like transformed two-dimension human face image.
4. the method according to claim 1, wherein the two-dimension human face area to the facial image to be identified
Area image carries out the outer registration process of plane
The two-dimension human face area image is generated into threedimensional model, projection function is determined according to the attitude angle, is based on the throwing
The threedimensional model of generation is projected to corresponding two-dimension human face image according to posture by shadow function.
5. the method according to claim 1, wherein the step 104 specifically includes:
Three-dimensional face data collection is obtained, the three-dimensional face model that the three-dimensional face data is concentrated is rotated to corresponding to be identified
Postrotational three-dimensional face model is projected to two dimensional image by the attitude angle of facial image, and according to corresponding characteristic point to throwing
The resulting two dimensional image of shadow does registration process in plane；And registration process outside plane, i.e. basis are done to the three-dimensional face model
Three-dimensional face model is projected to corresponding two-dimension human face image by the attitude angle, to generate the second multi-pose Face after alignment
Image set.
6. the method according to claim 1, wherein the convolutional neural networks be Inception-v4,
One of Inception-Resnet-v1, Inception-Resnet-v2.
7. the method according to claim 1, wherein by calculate vector between cosine similarity or it is European away from
From come the comparison that carries out the first averaged feature vector and the second averaged feature vector.
8. a kind of device based on three-dimensional face model identification two-dimension human face picture, which is characterized in that including at least one processing
Device, and the memory being connect at least one described processor communication；The memory be stored with can by it is described at least one
The instruction that processor executes, described instruction is executed by least one described processor, so that at least one described processor can
Method described in any one of perform claim requirement 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910082406.1A CN109858433B (en) | 2019-01-28 | 2019-01-28 | Method and device for identifying two-dimensional face picture based on three-dimensional face model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910082406.1A CN109858433B (en) | 2019-01-28 | 2019-01-28 | Method and device for identifying two-dimensional face picture based on three-dimensional face model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109858433A true CN109858433A (en) | 2019-06-07 |
| CN109858433B CN109858433B (en) | 2020-06-30 |
Family
ID=66896606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910082406.1A Active CN109858433B (en) | 2019-01-28 | 2019-01-28 | Method and device for identifying two-dimensional face picture based on three-dimensional face model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858433B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321821A (en) * | 2019-06-24 | 2019-10-11 | 深圳爱莫科技有限公司 | Face alignment initial method and device, storage medium based on tripleplane |
| CN112528902A (en) * | 2020-12-17 | 2021-03-19 | 四川大学 | Video monitoring dynamic face recognition method and device based on 3D face model |
| CN113808274A (en) * | 2021-09-24 | 2021-12-17 | 福建平潭瑞谦智能科技有限公司 | Face recognition model construction method and system and recognition method |
| CN117333928A (en) * | 2023-12-01 | 2024-01-02 | 深圳市宗匠科技有限公司 | Face feature point detection method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101021900A (en) * | 2007-03-15 | 2007-08-22 | 上海交通大学 | Method for making human face posture estimation utilizing dimension reduction method |
| CN105654048A (en) * | 2015-12-30 | 2016-06-08 | 四川川大智胜软件股份有限公司 | Multi-visual-angle face comparison method |
| CN107729875A (en) * | 2017-11-09 | 2018-02-23 | 上海快视信息技术有限公司 | Three-dimensional face identification method and device |
| CN108805977A (en) * | 2018-06-06 | 2018-11-13 | 浙江大学 | A kind of face three-dimensional rebuilding method based on end-to-end convolutional neural networks |
-
2019
- 2019-01-28 CN CN201910082406.1A patent/CN109858433B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101021900A (en) * | 2007-03-15 | 2007-08-22 | 上海交通大学 | Method for making human face posture estimation utilizing dimension reduction method |
| CN105654048A (en) * | 2015-12-30 | 2016-06-08 | 四川川大智胜软件股份有限公司 | Multi-visual-angle face comparison method |
| CN107729875A (en) * | 2017-11-09 | 2018-02-23 | 上海快视信息技术有限公司 | Three-dimensional face identification method and device |
| CN108805977A (en) * | 2018-06-06 | 2018-11-13 | 浙江大学 | A kind of face three-dimensional rebuilding method based on end-to-end convolutional neural networks |
Non-Patent Citations (2)
| Title |
|---|
| 谢宜江 等: "基于散斑立体匹配的快速三维人脸重建", 《光电子 激光》 * |
| 邹国锋 等: "多姿态人脸识别综述", 《模式识别与人工智能》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321821A (en) * | 2019-06-24 | 2019-10-11 | 深圳爱莫科技有限公司 | Face alignment initial method and device, storage medium based on tripleplane |
| CN110321821B (en) * | 2019-06-24 | 2022-10-25 | 深圳爱莫科技有限公司 | Human face alignment initialization method and device based on three-dimensional projection and storage medium |
| CN112528902A (en) * | 2020-12-17 | 2021-03-19 | 四川大学 | Video monitoring dynamic face recognition method and device based on 3D face model |
| CN113808274A (en) * | 2021-09-24 | 2021-12-17 | 福建平潭瑞谦智能科技有限公司 | Face recognition model construction method and system and recognition method |
| CN117333928A (en) * | 2023-12-01 | 2024-01-02 | 深圳市宗匠科技有限公司 | Face feature point detection method and device, electronic equipment and storage medium |
| CN117333928B (en) * | 2023-12-01 | 2024-03-22 | 深圳市宗匠科技有限公司 | Face feature point detection method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109858433B (en) | 2020-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| An et al. | Performance evaluation of model-based gait on multi-view very large population database with pose sequences | |
| Ge et al. | Robust 3d hand pose estimation in single depth images: from single-view cnn to multi-view cnns | |
| Guo et al. | Action recognition from video using feature covariance matrices | |
| Rahmani et al. | Histogram of oriented principal components for cross-view action recognition | |
| Wang et al. | Drpose3d: Depth ranking in 3d human pose estimation | |
| Papazov et al. | Real-time 3D head pose and facial landmark estimation from depth images using triangular surface patch features | |
| Sagonas et al. | 300 faces in-the-wild challenge: The first facial landmark localization challenge | |
| CN109858433A (en) | A kind of method and device based on three-dimensional face model identification two-dimension human face picture | |
| Ji et al. | Interactive body part contrast mining for human interaction recognition | |
| Lai et al. | A gesture-driven computer interface using Kinect | |
| Xu et al. | Human re-identification by matching compositional template with cluster sampling | |
| Li et al. | Morphable displacement field based image matching for face recognition across pose | |
| Holte et al. | Human action recognition using multiple views: a comparative perspective on recent developments | |
| Del Rincón et al. | Tracking human position and lower body parts using Kalman and particle filters constrained by human biomechanics | |
| Everingham et al. | Identifying individuals in video by combining'generative'and discriminative head models | |
| Li et al. | Expression-invariant face recognition with expression classification | |
| CN104050475A (en) | Reality augmenting system and method based on image feature matching | |
| Karaaba et al. | Robust face recognition by computing distances from multiple histograms of oriented gradients | |
| Han et al. | 3D face texture modeling from uncalibrated frontal and profile images | |
| Liu et al. | 3D action recognition using multiscale energy-based global ternary image | |
| Jampour et al. | Multi-view facial expressions recognition using local linear regression of sparse codes | |
| Shaikh et al. | Gait recognition using partial silhouette-based approach | |
| Nguyen et al. | Combined YOLOv5 and HRNet for high accuracy 2D keypoint and human pose estimation | |
| Pons-Moll et al. | Efficient and robust shape matching for model based human motion capture | |
| Sangineto et al. | Real-time viewpoint-invariant hand localization with cluttered backgrounds |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |