CN109522829A - A kind of smart phone " brush face " meeting register method based on deep learning - Google Patents
A kind of smart phone " brush face " meeting register method based on deep learning Download PDFInfo
- Publication number
- CN109522829A CN109522829A CN201811301566.2A CN201811301566A CN109522829A CN 109522829 A CN109522829 A CN 109522829A CN 201811301566 A CN201811301566 A CN 201811301566A CN 109522829 A CN109522829 A CN 109522829A
- Authority
- CN
- China
- Prior art keywords
- photo
- face
- meeting
- information
- deep learning
- 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
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
- G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C1/00—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people
- G07C1/10—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people together with the recording, indicating or registering of other data, e.g. of signs of identity
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Geometry (AREA)
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of smart phone " brush face " meeting register method based on deep learning, meeting initiator's initiating conference, conferenced party participates in meeting, participant and by the essential information of itself and photo upload into system, the essential information and photo are denoted as registration essential information and registration photo respectively；The public information for excavating scientific research personnel is acquired by web crawlers, public information includes essential information, photo and scientific research information, it is denoted as open essential information and open photo, using registration essential information, registration photo, open essential information, open photo as storage information；Will storage information store into MongoDB database as data comparison library；When identification, user inputs photo to be identified, photo to be identified is matched with the registration photo in data comparison library by the face identification method based on deep learning, if successful match, the storage information of the corresponding scientific research personnel of photo to be identified is pushed to user, to help the scientific research information of user acquisition scientific research personnel.
Description
Technical field
The invention belongs to computer visions and field of software development, since system is related to recognition of face, data acquisition is dug
The functional requirement in multiple entirely different fields such as pick, server of the system, system client, so used the technology of different field
It realizes.
Background technique
Face recognition technology is a research direction of biometric technology, with cost performance height, uniqueness, naturality
Good feature, therefore be concerned and be also widely used.Along with the development of deep learning, recognition of face starts to apply
In more and more scenes, it is related to the every aspect of life, such as Alipay brush face logs in, the unlock of mobile phone brush face, high-speed rail brush
Face enters the station.It is verified compared to traditional identity, does that authentication is more accurate using biological characteristic, reduce the possibility of fraud,
Increase the cost of fraud.The feature of face is had nothing in common with each other and relatively stable, therefore extracts that carry out authentication be highly effective
Method, but due to facial image will receive it is all multi-environment and it is biological itself variation influence so that recognition of face is ground
Studying carefully becomes incomparable complexity, and there are also various problems is urgently to be resolved at present.Face recognition technology, which is widely put into production, works as
In be unable to do without information technology, artificial intelligence, computer vision rapid development, so far recognition of face have become popular research neck
One of domain.
The frequent various academic conferences participated in both at home and abroad of scientific research personnel's meeting, but since personnel participating in the meeting is more, scientific research people
Member is often to the fewer of part plenary lecture people understanding, and speaker also makees detailed self-introduction without the time.Then, phase
Mutual not knowing about has led to the chance that every scientific research personnel has missed many exchanges.
Summary of the invention
Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provides a kind of intelligence based on deep learning
The public information of scientific research personnel is excavated in energy mobile phone " brush face " meeting register method, acquisition, and is stored into system database
As data comparison library, by the data of recognition of face Rapid matching scientific research personnel, so that help system user can be easily
The relevant informations such as research direction, the correlative theses treatise delivered and the project of hosting of acquisition speaker.
Technical solution: to achieve the above object, the technical solution adopted by the present invention are as follows:
A kind of smart phone " brush face " meeting register method based on deep learning, meeting initiator's initiating conference, meeting
Participant participates in meeting, participant and by the essential information of itself and photo upload into system, the essential information and photo point
It is not denoted as registration essential information and registration photo；The public information for excavating scientific research personnel, public information are acquired by web crawlers
Including essential information, photo and scientific research information, it is denoted as open essential information and open photo, registration essential information, registration are shone
Piece, open essential information, open photo are as storage information；Will storage information store into MongoDB database as data
Comparison database；When identification, user inputs photo to be identified, will be in photo to be identified and data comparison library by face identification method
Registration photo matched, if successful match, by the storage information of the corresponding scientific research personnel of photo to be identified to user
Push, to help the scientific research information of user acquisition scientific research personnel.
Further: further including that face registers and checks card, conferenced party enters link of registering, and the direction meeting is initiated in meeting
It discusses participant and sends the code of registering for being used for recognition of face, conferenced party shoots itself human face photo using client, by itself
Human face photo and code of registering upload to server-side, by itself human face photo and register photo progress by face identification method
Match, in successful match, and code of registering is registered under yard unanimous circumstances with what meeting initiator sent, success of registering.
Preferably, the face identification method the following steps are included:
Step 1, piece image is given, is realized using the cascade structure of a kind of combination conventional artificial feature and multi-layer perception (MLP)
Face datection；
Step 2, judge whether to detect face, if not having, register unsuccessfully；If performing the next step；
Step 3,5 key feature points, respectively two centers, nose and two are returned by cascading multiple depth models
Realize facial characteristics point location in the position of a corners of the mouth；
Step 4, convolutional neural networks model is improved, the feature vector set { v } for extracting image important area is special as face
Sign, and realize that face characteristic compares；Feature extraction with compare be recognition of face final step, its essence be calculate two width figures
Similarity score, before this be arranged a similarity threshold；The similarity calculating method of two width figures is as follows:
Wherein, viAnd vjIt is belonging respectively to the vector set { v } of two faces to be matched,Indicate viTransposition；Expression is looked into
Ask figure viWith vjχ2Distance, κ are freedom degree series,Indicate the incomplete Gamma function of low order, Γ (κ) is Gamma
Function.
The study of neural network model realizes that loss function is as follows by minimizing loss function L (θ, κ):
Wherein, SmlFor the similarity of two width figure m and l, SmmFor scheme m self-similarity,For regular terms, λ and θ are to want
The parameter of study；
Step 5, judge whether similarity is greater than given threshold, if more than successful match, otherwise it fails to match for label.
Preferred: face feature extraction method does not use traditional convolutional neural networks model, and it is direct to distinguish over master mould
It is important to extract image the present invention is based on the basis of facial characteristics point location for the convolutional neural networks feature for extracting entire image
The convolutional neural networks feature in region obtains feature vector set, and defines the effective side for calculating similarity and loss function
Method.
Preferred: MongoDB database represents database as a kind of of non-relational database, unlike relationship type number
According to library, does not need to specify and build table statement, the data of storage are stored in data file according to the form of class JSON key-value pair, have
The expansibility of height arbitrarily can record addition to a data and delete field without influencing other data records；Together
When the inquiry API that the provides relevant database inquiry mode that can match in excellence or beauty current, it might even be possible to established for some crucial key
Index；The data that acquisition is excavated can be quickly inserted into database.
It is preferred: acquisition excavate based on Python realize, use Beautiful Soup parsing library and
Request network request library.
The present invention compared with prior art, has the advantages that
1. the improved convolutional neural networks model of present invention use, the significantly more efficient important feature for excavating image,
It on the basis of this, proposes that effective image similarity calculates and loss function defines method, improves the timeliness and effectively of algorithm
Property.
It, can be for scholars even 2. the present invention can successfully obtain the personal homepage for the person of being queried by face alignment
Student enrollment provides unexpected convenience.Every scholar, which can pass through this platform and be easily accomplished, to hold a meeting, participates in meeting,
Moreover the conference management of this system is more combined with recognition of face, so that change of registering is light.Committee paper therein is total
It enjoys and has achieved the effect that participant's shared data, message is registered, and function meets participant's minutes, session discussing is demand.
3. the present invention not only includes the research and realization of face recognition algorithms and crawler technology, while including being based on Android system
The smart mobile phone application software development of system.Binding pattern identification technology, web data digging technology and applied software development of the present invention
Technology.
Detailed description of the invention
Fig. 1 meeting way system flow chart；
Fig. 2 recognition of face is registered flow chart；
The process flowchart of Fig. 3 acquisition user data；
Fig. 4 software function shows exemplary diagram.
Fig. 5 participant is registered exemplary diagram.
Fig. 6 software function shows exemplary diagram.
Fig. 7 meeting initiator code of registering generates and checks list exemplary diagram of registering.
Specific embodiment
In the following with reference to the drawings and specific embodiments, the present invention is furture elucidated, it should be understood that these examples are merely to illustrate this
It invents rather than limits the scope of the invention, after the present invention has been read, those skilled in the art are to of the invention various
The modification of equivalent form falls within the application range as defined in the appended claims.
A kind of smart phone " brush face " meeting Accreditation System based on deep learning, this system is based on realistic problem, solution
Certainly scholars in session when the research field that can not learn other side that is faced etc. problem.Using face recognition technology and data
Acquisition digging technology is developed, which mainly includes recognition of face, server of the system, data acquisition is excavated and client.
It also needs to undertake the logical demand of meeting in view of system not only needs to carry out recognition of face, is dug then relatively independent data acquire
The part of pick will be realized using independent subsystem.
System will further be subdivided into line module and meeting module on business module.
(1) recognition of face.Recognition of face includes three parts: Face datection, positioning feature point and face alignment, this three
A module is step necessary to completing a whole set of face identification system.
" brush face " meeting registration is realized in recognition of face.Recognition of face is the nucleus module of the system, comprising:
1. module can first verify whether file is legal after the facial image for taking client upload, only legal file
It can carry out the Face datection of next step.Business module saves file system of the image file in server of client upload
In, image file is read by locating file system, carries out the validity judgement of image file.
2. after the verifying by the first step, the detection of face is carried out to facial image for legal image.
3. carrying out the positioning of characteristic point after detecting face, current characteristic point is mainly 5 points, respectively two centers,
The position of nose and two corners of the mouths.
4. extracting the feature set of face, and compared, is obtained one by one with the picture library data in data acquisition digging system
The highest picture library data of similarity.
5. can first judge whether this similarity is greater than after face recognition module obtains the highest picture library data of similarity
Some specific threshold values, being only greater than the face that specific threshold values ability task two is opened on facial image is the same face,
So far the work of face recognition module is completed.Business module can only take the legal picture library data greater than threshold values.
(2) server of the system.The server-side of system is realized using the Express frame of Node.js.Server-side is from client
Receive the facial image that identifies of needs uploaded, take comparison result by the processing of recognition of face part, and then from database
The open source information of middle inquiry related scientific research personnel, finally returns to client for the open source information inquired.Since system can mention
For more functions, for example the function etc. that function, the member's content that face is registered are shared is provided, then server-side is at the beginning of design
It just needs, by being packaged Express, to realize that the skeleton of a simple multimode is answered in view of scalability
With.
(3) data acquisition is excavated.Data acquisition excavates part primarily to the scientific research personnel's data sent out in advance are adopted
Collection excavates, and acquires the data of excavation primarily to the acquisition for the scientific research personnel's data sent out in advance is excavated, including scientific research personnel
Facial image, direction of scientific rersearch and correlative theses etc..It acquires the data excavated and deposits the shared persistence number of Fang Yu server-side
According in the MongoDB of library.Acquisition excavates program and is mainly based upon Python realization, uses Beautiful Soup and parses library
And Request network request library.Capture program can initiate network request to specified Internet resources, upon receiving the response,
The data that acquisition is taken in parsing are carried out by Beautiful Soup parsing library.The data of acquisition can give MongoDB and carry out persistently
Change storage.
(4) client.Client is developed using the development scheme of Hybird, and thus client can be divided into two parts,
Android Native and Web H5, wherein Android Native carries out the H5 page by the Web Components View of Android
Encapsulation, is encapsulated in native applications.The main Web H5 that completes in the part Android Native is difficult to the calling camera completed
It takes pictures etc. and system bottom to be needed to provide the function of supporting, wherein the Web H5 page is based on Vue.js realization.User needs in user
Module can enter system after logging in, the personal information that line module maintains user includes personal brief introduction, individual research
Achievement etc..User can also check other people public information in systems.The userspersonal information of this part is to pass through number mostly
It is excavated according to acquisition digging system acquisition, does not need user oneself and go to safeguard the personal information of oneself, user is in system
In only need to be concerned about and participate in meeting, creation meeting etc..
It is as shown in Figure 1 meeting way system flow chart of the present invention, the realization of entire meeting way system includes logging in, sending out
Meeting, shared file are played, meeting, recognition of face is added endorses to code and the functions such as register.
Wherein, as meeting initiator, passing through the logical app of meeting may be implemented following functions:
1) initiating conference and the information such as meeting-place, time are selected；
2) committee paper is uploaded, is checked for participant；
3) list of registering of participant is checked；
4) information of participant is checked；
5) leaving meeting message is participated in interact with everybody；
As the participant of meeting, pass through that meeting is logical to may be implemented following functions:
6) list of conference currently participated in is checked；
7) it clicks and participates in meeting, and check meeting shared file；
8) it is registered and is checked card using face；
9) message and everybody council related content are participated in.
The realization of entire meeting way system is endorsed including login, initiating conference, shared file, addition meeting, recognition of face
It the functions such as registers to code.
A kind of smart phone " brush face " meeting register method based on deep learning, meeting initiator's initiating conference, meeting
Participant participates in meeting, participant and by the essential information of itself and photo upload into system, the essential information and photo point
It is not denoted as registration essential information and registration photo；The public information for excavating scientific research personnel, public information are acquired by web crawlers
Including essential information, photo and scientific research information, it is denoted as open essential information and open photo, registration essential information, registration are shone
Piece, open essential information, open photo are as storage information；Will storage information store into MongoDB database as data
Comparison database；When identification, user inputs photo to be identified, will be in photo to be identified and data comparison library by face identification method
Registration photo matched, if successful match, by the storage information of the corresponding scientific research personnel of photo to be identified to user
Push, to help the scientific research information of user acquisition scientific research personnel.
As shown in Fig. 2, face recognition process people of the invention is broadly divided into three parts: Face datection, face feature point are fixed
Position, Characteristic Contrast and extraction, comprising the following steps:
1) piece image is given, face is realized using the cascade structure of a kind of combination conventional artificial feature and multi-layer perception (MLP)
Detection；
2) judge whether to detect face, if not having, register unsuccessfully；If performing the next step；
3) pass through and cascade multiple depth model (stack autoencoder network) 5 key feature points of Lai Huigui, respectively two
Realize facial characteristics point location in the position at center, nose and two corners of the mouths；
4) convolutional neural networks model is improved, the feature vector set { v } for extracting image important area is used as face characteristic, and
Realize that face characteristic compares；The similarity calculating method of two width figures is as follows:
Wherein, viAnd vjIt is belonging respectively to the vector set { v } of two faces to be matched,Indicate viTransposition；Expression is looked into
Ask figure viWith vjχ2Distance, κ are freedom degree series,Indicate the incomplete Gamma function of low order, Γ (κ) is Gamma
Function.
The study of neural network model realizes that loss function is as follows by minimizing loss function L (θ, κ):
Wherein, SmlFor the similarity of two width figure m and l, SmmFor scheme m self-similarity,For regular terms, λ and θ are to want
The parameter of study；
Feature extraction with compare be recognition of face final step, its essence be calculate two width figures similarity score,
We will be arranged a threshold value before this, this threshold value is by largely testing obtaining as a result, not being for the same person and not
With the line of demarcation of people.It compares in two steps, one is the registration phase of original picture library face, likens adult first meeting to the most
Properly, second step is to work as system to meet again with this people, judges whether it is the stage of same people.
5) judge whether likeness in form degree is greater than given threshold, if more than success of registering, otherwise register unsuccessfully.
It is illustrated in figure 3 the process flowchart of acquisition user data of the invention, system is excavated in independent data acquisition
The public information of scientific research personnel can be acquired and be excavated into system by system.Complete the process of data acquisition are as follows:
1) acquisition is excavated program and is acquired to the personal public information of scientific research personnel, and the data of acquisition include scientific research personnel
Essential information, personal brief introduction, personal achievement etc., wherein the facial image of scientific research personnel can acquire together with essential information
It arrives.
2) data that storage acquisition is excavated are into MongoDB, a kind of representative of the MongoDB as non-relational database
Database, unlike relevant database, do not need it is specified build table statement, the data of storage according to class JSON key-value pair form
Be stored in data file, have the expansibility of height, can arbitrarily to a data record addition delete field without
Influence other data records.The relevant database inquiry mode that the inquiry API provided simultaneously can match in excellence or beauty current, it is succinct and strong
Greatly, it might even be possible to establish index for some crucial key.In view of the powerful of MongoDB database, we can be very easily
The data that acquisition is excavated quickly are inserted into database.
Exemplary diagram is shown for software function of the invention as shown in figs. 4-7, is specifically included that
1) user logs into system in line module.
2) after user logs into system, in homepage just it can be seen that nearest list of conference and Client-initiated meeting
With the list of conference of participation.Clicking any meeting in list just can check the specific content of meeting, including meeting title,
The time of meeting, meeting-place and shared committee paper etc..
3) if it is intended to participating in some meeting, creation meeting can be clicked, inputs meeting relevant information and personal authentication's letter
Breath.
4) registering for meeting, which not only can be generated, in the founder of meeting code but also can check current meeting signature people
Member, clicking lists of persons of registering can see the personal information of corresponding personnel participating in the meeting.
5) if it is intended to participating in some meeting, the button that participation meeting is clicked in the details page of the meeting can be identified oneself with
In meeting.Enter if having had been participating in meeting and having clicked button and register link, what input conference creation person provided register code into
Row recognition of face is registered.It needs to take pictures using client when registering and uploads to the verifying that server-side carries out recognition of face, only
It just registers at last success in the case where face verification and code verifying of registering are adopted unanimously.
Server-side receive register request when, can first take photo resources, the recognition of face part of calling system carries out
Photo of registering carries out identification with picture library photo and compares, and returns to the picture library data for being greater than specific threshold.Server-side is collecting
Data in search the picture library data, otherwise success of registering if picture library data are consistent with user data of registering is registered unsuccessfully.It is logical
The mode for crossing recognition of face, which register, can effectively avoid the case where attending a meeting instead of other people.
The present invention is as meeting initiator, and passing through the logical app of meeting may be implemented following functions: initiating conference simultaneously selects meeting
The information such as place, time；Committee paper is uploaded, is checked for participant；Check the list of registering of participant；Check the letter of participant
Breath；Leaving meeting message is participated in interact with everybody.As the participant of meeting, pass through that meeting is logical to may be implemented following functions: checking mesh
The list of conference of preceding participation；It clicks and participates in meeting, and check meeting shared file；It is registered using face；Participate in message and everybody
Council related content.The present invention is an application of recognition of face, is to be embodied in the form of app, wherein being applied to client
End further comprises Android and web and is embedded in two aspects, and bottom realizes that rear end is realized using nodejs based on C++, and data are logical
Mongodb database access is crossed, using fused data excavation and access method, is realized fast and accurately using deep learning method
Recognition of face and certification.
The above is only a preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (6)
1. a kind of smart phone " brush face " meeting register method based on deep learning, it is characterised in that: meeting initiator initiates
Meeting, conferenced party participate in meeting, participant and by the essential information of itself and photo upload into system, the essential information
It is denoted as registration essential information and registration photo respectively with photo；The public information for excavating scientific research personnel is acquired by web crawlers,
Public information includes essential information, photo and scientific research information, is denoted as open essential information and open photo, by the basic letter of registration
Breath, registration photo, open essential information, open photo are as storage information；Storage information is stored into MongoDB database
As data comparison library；When identification, user inputs photo to be identified, will be to by the face identification method based on deep learning
Identification photo is matched with the registration photo in data comparison library, if successful match, by the corresponding scientific research of photo to be identified
The storage information of personnel is pushed to user, to help the scientific research information of user acquisition scientific research personnel.
2. the smart phone based on deep learning " brush face " meeting register method according to claim 1, it is characterised in that: also
It registers and checks card including face, conferenced party enters link of registering, and meeting is initiated the direction conferenced party and sent for people
The code of registering of face identification, conferenced party shoot itself human face photo using client, by itself human face photo and register on code
Server-side is passed to, is matched itself human face photo with registration photo by the face identification method based on deep learning,
In successful match, and code of registering is registered under yard unanimous circumstances with what meeting initiator sent, success of registering.
3. the smart phone based on deep learning " brush face " meeting register method according to claim 1, which is characterized in that institute
State the face identification method based on deep learning the following steps are included:
Step 1, piece image is given, face is realized using the cascade structure of a kind of combination conventional artificial feature and multi-layer perception (MLP)
Detection；
Step 2, judge whether to detect face, if not having, register unsuccessfully；If performing the next step；
Step 3,5 key feature points, respectively two centers, nose and two mouths are returned by cascading multiple depth models
Realize facial characteristics point location in the position at angle；
Step 4, convolutional neural networks model is improved, the feature vector set { v } for extracting image important area is used as face characteristic, and
Realize that face characteristic compares；Feature extraction with compare be recognition of face final step, its essence be calculate two width figures phase
Like degree score, a similarity threshold is set before this；The similarity calculating method of two width figures is as follows:
Wherein, viAnd vjIt is belonging respectively to the vector set { v } of two faces to be matched,Indicate viTransposition；Indicate query graph vi
With vjχ2Distance, κ are freedom degree series,Indicate the incomplete Gamma function of low order, Γ (κ) is Gamma function.
The study of neural network model realizes that loss function is as follows by minimizing loss function L (θ, κ):
Wherein, SmlFor the similarity of two width figure m and l, SmmFor scheme m self-similarity,For regular terms, λ and θ are to learn
Parameter；
Step 5, judge whether similarity is greater than given threshold, if more than successful match, otherwise it fails to match for label.
4. the smart phone based on deep learning " brush face " meeting register method according to claim 1, it is characterised in that: people
Face characteristic extracting method does not use traditional convolutional neural networks model, distinguishes over the convolution that master mould directly extracts entire image
Neural network characteristics extract the convolutional neural networks of image important area the present invention is based on the basis of facial characteristics point location
Feature obtains feature vector set, and defines the effective method for calculating similarity and loss function.
5. the smart phone based on deep learning " brush face " meeting register method according to claim 1, it is characterised in that:
MongoDB database represents database as a kind of of non-relational database, unlike relevant database, does not need to specify
Table statement is built, the data of storage are stored in data file according to the form of class JSON key-value pair, have the expansibility of height,
Addition arbitrarily can be recorded on a data and delete field without influencing other data records；The inquiry API provided simultaneously
The relevant database inquiry mode that can match in excellence or beauty current, it might even be possible to establish index for some crucial key；It can will acquire
The data of excavation are quickly inserted into database.
6. the smart phone based on deep learning " brush face " meeting register method according to claim 1, it is characterised in that: adopt
Collection is excavated to be realized based on Python, uses the parsing library Beautiful Soup and Request network request library.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811301566.2A CN109522829B (en) | 2018-11-02 | 2018-11-02 | Smart phone face-brushing conference registration method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811301566.2A CN109522829B (en) | 2018-11-02 | 2018-11-02 | Smart phone face-brushing conference registration method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109522829A true CN109522829A (en) | 2019-03-26 |
| CN109522829B CN109522829B (en) | 2021-10-26 |
Family
ID=65774158
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811301566.2A Active CN109522829B (en) | 2018-11-02 | 2018-11-02 | Smart phone face-brushing conference registration method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109522829B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414798A (en) * | 2019-07-03 | 2019-11-05 | 天津市多智信息科技有限公司 | A kind of meeting signature system and method based on video human face identification |
| CN111144292A (en) * | 2019-12-26 | 2020-05-12 | 武汉兴图新科电子股份有限公司 | System for controlling account of audio and video platform based on face recognition technology |
| CN111368766A (en) * | 2020-03-09 | 2020-07-03 | 云南安华防灾减灾科技有限责任公司 | Cattle face detection and identification method based on deep learning |
| CN111899135A (en) * | 2020-07-04 | 2020-11-06 | 深圳市联想空间艺术工程有限公司 | Intelligent companion chemical method and system based on face recognition |
| WO2021179706A1 (en) * | 2020-03-13 | 2021-09-16 | 平安科技（深圳）有限公司 | Meeting check-in method and system, computer device, and computer readable storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6526168B1 (en) * | 1998-03-19 | 2003-02-25 | The Regents Of The University Of California | Visual neural classifier |
| CN104537341A (en) * | 2014-12-23 | 2015-04-22 | 北京奇虎科技有限公司 | Human face picture information obtaining method and device |
| US20160373437A1 (en) * | 2015-02-15 | 2016-12-22 | Beijing Kuangshi Technology Co., Ltd. | Method and system for authenticating liveness face, and computer program product thereof |
| CN106503669A (en) * | 2016-11-02 | 2017-03-15 | 重庆中科云丛科技有限公司 | A kind of based on the training of multitask deep learning network, recognition methods and system |
| CN107292986A (en) * | 2017-07-11 | 2017-10-24 | 北京眼神科技有限公司 | A kind of conference service method, server and host terminal |
| CN108090981A (en) * | 2017-11-17 | 2018-05-29 | 克立司帝控制系统（上海）有限公司 | Meeting signature system and method based on face recognition technology |
| CN108090223A (en) * | 2018-01-05 | 2018-05-29 | 牛海波 | A kind of opening scholar portrait method based on internet information |
-
2018
- 2018-11-02 CN CN201811301566.2A patent/CN109522829B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6526168B1 (en) * | 1998-03-19 | 2003-02-25 | The Regents Of The University Of California | Visual neural classifier |
| CN104537341A (en) * | 2014-12-23 | 2015-04-22 | 北京奇虎科技有限公司 | Human face picture information obtaining method and device |
| US20160373437A1 (en) * | 2015-02-15 | 2016-12-22 | Beijing Kuangshi Technology Co., Ltd. | Method and system for authenticating liveness face, and computer program product thereof |
| CN106503669A (en) * | 2016-11-02 | 2017-03-15 | 重庆中科云丛科技有限公司 | A kind of based on the training of multitask deep learning network, recognition methods and system |
| CN107292986A (en) * | 2017-07-11 | 2017-10-24 | 北京眼神科技有限公司 | A kind of conference service method, server and host terminal |
| CN108090981A (en) * | 2017-11-17 | 2018-05-29 | 克立司帝控制系统（上海）有限公司 | Meeting signature system and method based on face recognition technology |
| CN108090223A (en) * | 2018-01-05 | 2018-05-29 | 牛海波 | A kind of opening scholar portrait method based on internet information |
Non-Patent Citations (2)
| Title |
|---|
| JIWHAN KIM ET AL: "3D face recognition via discriminative keypoint selection", 《2017 14TH INTERNATIONAL CONFERENCE ON UBIQUITOUS ROBOTS AND AMBIENT INTELLIGENCE》 * |
| 周鑫焱等: "融合局部方向模式和卷积神经网络的人脸识别", 《计算机工程与设计》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414798A (en) * | 2019-07-03 | 2019-11-05 | 天津市多智信息科技有限公司 | A kind of meeting signature system and method based on video human face identification |
| CN111144292A (en) * | 2019-12-26 | 2020-05-12 | 武汉兴图新科电子股份有限公司 | System for controlling account of audio and video platform based on face recognition technology |
| CN111368766A (en) * | 2020-03-09 | 2020-07-03 | 云南安华防灾减灾科技有限责任公司 | Cattle face detection and identification method based on deep learning |
| CN111368766B (en) * | 2020-03-09 | 2023-08-18 | 云南安华防灾减灾科技有限责任公司 | Deep learning-based cow face detection and recognition method |
| WO2021179706A1 (en) * | 2020-03-13 | 2021-09-16 | 平安科技（深圳）有限公司 | Meeting check-in method and system, computer device, and computer readable storage medium |
| CN111899135A (en) * | 2020-07-04 | 2020-11-06 | 深圳市联想空间艺术工程有限公司 | Intelligent companion chemical method and system based on face recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109522829B (en) | 2021-10-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109522829A (en) | A kind of smart phone " brush face " meeting register method based on deep learning | |
| CN110414373B (en) | Deep learning palm vein recognition system and method based on cloud edge-side cooperative computing | |
| CN108764065A (en) | A kind of method of pedestrian's weight identification feature fusion assisted learning | |
| Vascon et al. | Detecting conversational groups in images and sequences: A robust game-theoretic approach | |
| CN107025397B (en) | Identity information acquisition method and device | |
| CN106803289A (en) | A kind of false proof method and system of registering of intelligent mobile | |
| CN106203372A (en) | Eye-based living body detection method and device and terminal equipment | |
| CN110555651A (en) | Intelligent express processing method | |
| CN107944379A (en) | White of the eye image super-resolution rebuilding and image enchancing method based on deep learning | |
| CN104167025A (en) | Mobile attendance system based on dynamic face recognition and WIFI positioning and method of attendance system | |
| WO2020244174A1 (en) | Face recognition method, apparatus and device, and computer readable storage medium | |
| WO2021179706A1 (en) | Meeting check-in method and system, computer device, and computer readable storage medium | |
| CN112801054B (en) | Face recognition model processing method, face recognition method and device | |
| CN106303599A (en) | A kind of information processing method, system and server | |
| CN106383913A (en) | Face brushing friend-making method and device based on communication terminal | |
| CN110033399A (en) | A kind of information publication, extracting method and the system of the wisdom class board based on recognition of face | |
| CN105184267A (en) | Face-identification-based secondary-deformation auxiliary authorization method | |
| CN109272621A (en) | A kind of management system and method based on face recognition technology | |
| CN113111330A (en) | Enterprise legal identity authentication management method and system based on enterprise chain code | |
| Sivakumar et al. | Iot based intelligent attendance monitoring with face recognition scheme | |
| Shao et al. | Federated test-time adaptive face presentation attack detection with dual-phase privacy preservation | |
| Ishwarya et al. | Detection of face mask using convolutional neural network | |
| Geetha et al. | Implementation of a Blockchain Based Attendance Tracking System | |
| CN102930616A (en) | Smiling face sign-in system for early childhood education | |
| KR102642001B1 (en) | Metaverse-based video chat service system and method |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20190326 Assignee: NUPT INSTITUTE OF BIG DATA RESEARCH AT YANCHENG Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2021980013920 Denomination of invention: A smart phone "face brushing" conference registration method based on deep learning Granted publication date: 20211026 License type: Common License Record date: 20211202 |