CN106803301A - A kind of recognition of face guard method and system based on deep learning - Google Patents
A kind of recognition of face guard method and system based on deep learning Download PDFInfo
- Publication number
- CN106803301A CN106803301A CN201710192816.2A CN201710192816A CN106803301A CN 106803301 A CN106803301 A CN 106803301A CN 201710192816 A CN201710192816 A CN 201710192816A CN 106803301 A CN106803301 A CN 106803301A
- Authority
- CN
- China
- Prior art keywords
- face
- image information
- gate inhibition
- recognition
- described 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
- 238000013135 deep learning Methods 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 52
- 230000005764 inhibitory process Effects 0.000 claims abstract description 91
- 238000012545 processing Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 5
- 230000005540 biological transmission Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 description 19
- 238000012549 training Methods 0.000 description 12
- 230000008030 elimination Effects 0.000 description 8
- 238000003379 elimination reaction Methods 0.000 description 8
- 230000004927 fusion Effects 0.000 description 8
- 238000013527 convolutional neural network Methods 0.000 description 7
- 238000012795 verification Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 230000007935 neutral effect Effects 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000011478 gradient descent method Methods 0.000 description 3
- 230000006698 induction Effects 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000001965 increasing effect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000035515 penetration Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 235000019504 cigarettes Nutrition 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000005059 dormancy Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000003595 mist Substances 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000035807 sensation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000012772 sequence design Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000010025 steaming Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- 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
- G07C9/00—Individual registration on entry or exit
- G07C9/20—Individual registration on entry or exit involving the use of a pass
- G07C9/22—Individual registration on entry or exit involving the use of a pass in combination with an identity check of the pass holder
- G07C9/25—Individual registration on entry or exit involving the use of a pass in combination with an identity check of the pass holder using biometric data, e.g. fingerprints, iris scans or voice recognition
-
- 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
- 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
- G07C9/00—Individual registration on entry or exit
- G07C9/30—Individual registration on entry or exit not involving the use of a pass
- G07C9/32—Individual registration on entry or exit not involving the use of a pass in combination with an identity check
- G07C9/37—Individual registration on entry or exit not involving the use of a pass in combination with an identity check using biometric data, e.g. fingerprints, iris scans or voice recognition
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of recognition of face guard method based on deep learning and system, wherein the method includes:Receive gate inhibition and eliminate request, judge that the gate inhibition eliminates whether request is sent by designated terminal, if it is not, then determining that the gate inhibition eliminates request and sent by IMAQ terminal;Obtain the collection of described image acquisition terminal eliminates the corresponding image information of request with the gate inhibition, and recognition of face is carried out to described image information using deep learning face recognition algorithms, obtains corresponding face recognition result;Judge the whether corresponding default face of the face recognition result, if it is, indicate gate control system to eliminate gate inhibition, if it is not, then refusal eliminates gate inhibition.Because face can uniquely identify a people and be difficult to forge, therefore the control to gate control system is realized by recognition of face in the application, substantially increase the security of gate control system.
Description
Technical field
The present invention relates to technical field of electronic equipment, more specifically to a kind of recognition of face based on deep learning
Guard method and system.
Background technology
With the improvement of people ' s living standards, people more focus on the safety of domestic environment, and security protection idea is constantly strengthened;Companion
With the raising of this demand, intelligent access control system arises at the historic moment, and increasing enterprise, retail shop, family are assembled with various
The gate control system of various kinds.
The current commonplace gate control system for using is nothing more than video gate inhibition, password access, radio frequency gate inhibition or fingerprint access control
Etc..Wherein, video gate inhibition is that video information is simply sent to user, and without intellectuality, is substantially be unable to do without
" people's air defense ", can not definitely ensure house security when user is absent from the scene;The maximum hard defects of password access are that password is easily forgotten,
And easily crack;The shortcoming of radio frequency gate inhibition is then " recognizing card not recognize people ", and radio-frequency card is easily lost and is easily usurped by other people;Separately
Outward, the potential safety hazard of fingerprint access control is then that fingerprint is easily replicated.Therefore, the above-mentioned gate control system for providing in the prior art is corresponded to
Reason has that security is relatively low.
In sum, how to provide a kind of security gate control system correspondence technical scheme higher, be current this area skill
Art personnel's problem demanding prompt solution.
The content of the invention
It is an object of the invention to provide a kind of recognition of face guard method based on deep learning and system, to ensure gate inhibition
System has security higher.
To achieve these goals, the present invention provides following technical scheme:
A kind of recognition of face guard method based on deep learning, including:
Receive gate inhibition and eliminate request, judge that the gate inhibition eliminates whether request is sent by designated terminal, if it is not, then determining
The gate inhibition eliminates request and is sent by IMAQ terminal;
Obtain the collection of described image acquisition terminal eliminates the corresponding image information of request with the gate inhibition, using depth
Practise face recognition algorithms carries out recognition of face to described image information, obtains corresponding face recognition result;
Judge the whether corresponding default face of the face recognition result, if it is, indicate gate control system to eliminate gate inhibition, such as
Really no, then refusal eliminates gate inhibition.
Preferably, obtain the collection of described image acquisition terminal with the gate inhibition eliminate the corresponding image information of request it
Afterwards, also include:
Described image information is identified using deep learning photo array algorithm, if identifying described image information
Real face is shot, is then performed the utilization deep learning face recognition algorithms to described image information
The step of carrying out recognition of face, if identifying that described image information is that the face of photo is carried out shooting what is obtained, refuses
Recognition of face is carried out to described image information.
Preferably, also include:
If it is that the face of photo is carried out that the face recognition result does not correspond to default face or described image information
Shooting is obtained, then send and carry the warning information of the face recognition result or described image information to specified end
End.
Preferably, send and carry the warning information of the face recognition result or described image information to the specified end
After end, also include:
Obtain the designated terminal and receive the command information returned after the warning information, perform the command information simultaneously
The command information and corresponding face recognition result or image information are stored, to detect the people of storage again
Corresponding command information is performed when face recognition result or described image information.
Preferably, also include:
If the face recognition result does not correspond to default face or is that the face of photo is carried out shooting what is obtained,
Outwardly show the information of authentication failed.
Preferably, also include:
If the gate inhibition eliminates request and is sent by the designated terminal, it indicates that the gate control system eliminates door
Prohibit.
Preferably, also include:
Judge whether that someone enters in designated area using human body infrared inductor, if it is, indicating described image to adopt
Collection terminal is into normal mode of operation and carries out the collection of image information, if it is not, then indicating described image acquisition terminal to keep
Acquiescence park mode set in advance.
Preferably, obtain the collection of described image acquisition terminal with the gate inhibition eliminate the corresponding image information of request it
Afterwards, also include:
The ccd image information and Infrared Image Information that will be included in described image information are merged, and perform the utilization
The step of deep learning face recognition algorithms carry out recognition of face to described image information.
Preferably, recognition of face is carried out to described image information using deep learning face recognition algorithms, including:
Recognition of face is carried out to described image information using the deep learning face recognition algorithms realized based on GPU.
A kind of face recognition door control system based on deep learning, including:
First judge module, is used for:Receive gate inhibition and eliminate request, judge that whether the gate inhibition eliminates request by designated terminal
Send, if it is not, then determining that the gate inhibition eliminates request and sent by IMAQ terminal;
Image processing module, is used for:Obtain the corresponding with gate inhibition elimination request of described image acquisition terminal collection
Image information, recognition of face is carried out using deep learning face recognition algorithms to described image information, is obtained corresponding face and is known
Other result;
Second judge module, is used for:Whether the corresponding default face of the face recognition result is judged, if it is, indicating
Gate control system eliminates gate inhibition, if it is not, then refusal eliminates gate inhibition.
The invention provides a kind of recognition of face guard method based on deep learning and system, wherein the method includes:
Receive gate inhibition and eliminate request, judge that the gate inhibition eliminates whether request is sent by designated terminal, if it is not, then determining the gate inhibition
Request is eliminated to be sent by IMAQ terminal;Obtain the corresponding with gate inhibition elimination request of described image acquisition terminal collection
Image information, recognition of face is carried out using deep learning face recognition algorithms to described image information, is obtained corresponding face and is known
Other result;Judge the whether corresponding default face of the face recognition result, if it is, indicate gate control system to eliminate gate inhibition, such as
Really no, then refusal eliminates gate inhibition.When gate inhibition eliminates request sent by IMAQ terminal in technical characteristic disclosed in the present application
When, the corresponding image information of system is eliminated to the gate inhibition carries out the identification of deep learning face recognition algorithms, so as to judge figure
As the whether corresponding default face of the face recognition result of information, if it is, the owner of explanation image information correspondence face has
The authority of gate inhibition is eliminated, now indicates gate control system to eliminate gate inhibition, otherwise, then illustrate that the owner of image information correspondence face does not have
There is the authority for eliminating gate inhibition, now refusal eliminates gate inhibition, because face can uniquely identify a people and be difficult to forge, therefore
The control to gate control system is realized by recognition of face in the application, the security of gate control system is substantially increased.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Inventive embodiment, for those of ordinary skill in the art, on the premise of not paying creative work, can also basis
The accompanying drawing of offer obtains other accompanying drawings.
Fig. 1 is a kind of flow chart of recognition of face guard method based on deep learning provided in an embodiment of the present invention;
Fig. 2 is deep learning people in a kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention
The algorithm structure figure of face recognizer;
Fig. 3 is deep learning people in a kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention
The illustraton of model of face recognizer;
Fig. 4 is image co-registration calculation in a kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention
The Organization Chart of method;
Fig. 5 is image co-registration calculation in a kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention
The flow chart of method;
Fig. 6 is a kind of structural representation of face recognition door control system based on deep learning provided in an embodiment of the present invention
Figure.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
Fig. 1 is referred to, it illustrates a kind of recognition of face gate inhibition side based on deep learning provided in an embodiment of the present invention
The flow chart of method, can include:
S11:Receive gate inhibition and eliminate request, judge that gate inhibition eliminates whether request is sent by designated terminal, if it is not, then determining
Gate inhibition eliminates request and is sent by IMAQ terminal.
Wherein designated terminal generally corresponds to the owner of gate control system, or other set in advance have to gate control system
The terminal of the people of control authority, if gate inhibition eliminates system to be sent by designated terminal, explanatory diagram is adopted as acquisition terminal
The image information of visiting personnel is collected and then have sent gate inhibition and eliminated request, now needed to obtain the image information of collection and sentence
Whether the visiting personnel of the image information of breaking correspondence can realize the control to gate control system.
S12:Obtain the collection of IMAQ terminal eliminates the corresponding image information of request with gate inhibition, using deep learning people
Face recognizer carries out recognition of face to image information, obtains corresponding face recognition result.
S13:Judge the whether corresponding default face of face recognition result, if it is, indicate gate control system to eliminate gate inhibition, such as
Really no, then refusal eliminates gate inhibition.
Recognition of face is carried out to image information using deep learning face recognition algorithms, if the recognition of face that identification is obtained
Result shows the visiting default face of personnel's correspondence, then illustrate that visiting personnel have the authority of access control system, now indicate door
Access control system eliminates gate inhibition, namely indicates gate control system to open door lock, it is allowed in visiting personnel enter, and otherwise, then illustrates visiting personnel
Authority without access control system, now refusal elimination gate inhibition.It is that gate control system is possessed of control power wherein to preset face
The face information of the people of limit.
It is when being sent by IMAQ terminal, to the gate inhibition when gate inhibition eliminates request in technical characteristic disclosed in the present application
The corresponding image information of elimination system carries out the identification of deep learning face recognition algorithms, so as to judge the face of image information
The whether corresponding default face of recognition result, if it is, the owner of explanation image information correspondence face has the power for eliminating gate inhibition
Limit, now instruction gate control system elimination gate inhibition, otherwise, then illustrates the owner of image information correspondence face without elimination gate inhibition's
Authority, now refusal elimination gate inhibition because face can uniquely identify a people and be difficult to forge, therefore passes through in the application
The control to gate control system is realized in recognition of face, substantially increases the security of gate control system.
Additionally, technical scheme disclosed in the present application is using current state-of-the-art based on the realization of deep learning face recognition algorithms
The identification of image information, compared with currently used more algorithm such as PCA, SVM, LBP, identification of the deep learning to identity characteristic
Accuracy rate is higher, has been even more than human eye discrimination, therefore technical scheme disclosed in the present application also has recognition of face accurate
The characteristics of really rate is high, further increases the security of gate control system.
It is further to note that in technical scheme disclosed in the present application, face verification is laid particular emphasis on rather than recognition of face,
Reduce class inherited such that it is able to effective, it is easy to expand to other application, and integration across database is effective;When in data block
When classification is more, its generalization ability is also stronger.Specifically, the field of application concerns is the subdomains of recognition of face ---
Face verification, be exactly in simple terms judge two pictures whether same person.So, face verification problem is easy for
Recognition of face problem can be changed into, recognition of face is exactly to carry out multiple face verification.Learnt to one using deep learning method
The set that group high dimensional feature is represented is used for face verification, is then learnt by carrying out a recognition of face task for multicategory classification
Feature, and the extensive new authentications not recognized with other to face verification of feature.Identity characteristic takes from last
The activation value of individual hidden layer.Meanwhile, multicategory classification is carried out to all of identity, rather than two classification, this is examined based on two
Consider:One is, a training sample is trained to the class in multiple classes, more difficult than carrying out two classification, this challenge energy
The super learning ability of neutral net is enough made full use of to extract validity feature;Two are, impliedly increase on convolutional neural networks
Strong regularization is added, has helped to create and effective shared hidden layer of classifying is represented.Therefore, the feature for learning has general well
Change ability.The above-mentioned algorithm framework that the present invention is provided can be as shown in Figure 2.
The identification division algorithm of the application is mainly made up of depth convolutional neural networks and identification and classification device, specifically can be as
Shown in Fig. 3, wherein model parameter is as follows:
Ground floor convolutional layer:Convolution kernel size 4 × 4, port number is 3;Output characteristic figure size is 36 × 36, and totally 20 lead to
Road.
Ground floor pond layer:Core size is 2 × 2;Output sampled images size is 18 × 18, totally 20 passages.
Second layer convolutional layer:Convolution kernel size 3 × 3, port number is 20;Output characteristic figure size is 16 × 16, totally 40
Passage.
Second layer pond layer:Core size is 2 × 2;Output sampled images size is 8 × 8, totally 40 passages.
Third layer convolutional layer:Convolution kernel size 3 × 3, port number is 40;Output characteristic figure size is 6 × 6, and totally 80 lead to
Road.
Third layer pond layer:Core size is 2 × 2;Output sampled images size is 3 × 3, totally 80 passages.
First full articulamentum:Using Maxout activation primitives, 160 dimensional vectors are exported.
Second full articulamentum:Using Maxout activation primitives, 160 dimensional vectors are exported.
The RGB Three Channel Color facial images that the mode input is one 39 × 39 × 3, first pass around ground floor convolutional layer and enter
Row feature extraction.Convolutional layer extract characteristic pattern formula be:
fij=sigmoid ((W*x)ij+b)
Above-mentioned formula means that the i row j row pixels of characteristic pattern are the identical bits of each passage by convolution kernel Yu input picture
The convolution results put are added and take activation value again.Wherein, W is the weight parameter of neutral net, and b is bias term parameter, activation primitive
It is sigmoid (z)=1/ (1+e-z).Local convolution operation is easier to perceive local feature compared to full connection, especially to face
Face feature, can sensitively extract, and weight parameter can be greatly reduced.But in this way, weight parameter still mistake
It is many, over-fitting is susceptible to, learning characteristic is not easy to, it is necessary to further reduce parameter, pond is then input into after convolutional layer
Layer (also referred to as sample level).Pond layer means that the regional area in characteristic pattern uses same parameter, can effectively reduce parameter.Here
Using maximum pond method, sampling formula is as follows:
Wherein pijIt is the output image of Chi Huahou, xijIt is input picture, the max function purpose in above formula is to ask pond
Max pixel value point in the range of core, the pixel in same pond region shares same weight parameter.So not only achieve
The feature of more low dimensional, and the problem that over-fitting occurs can be avoided.
The wherein more general neutral net of convolutional network has the following advantages in terms of image procossing:
The topological structure of input picture and network can preferably coincide.
Feature extraction and pattern classification are carried out simultaneously, and the generation in training simultaneously.
The shared training parameter that can reduce network of weight, makes neural network structure become simpler, and adaptability is stronger.
Compared to traditional convolutional neural networks, the present invention instead of original pure convolutional layer using multilayer perceptron+convolution.
Because convolution is linear operation, it is not easy to learn nonlinear characteristic, and multilayer perceptron learning of nonlinear functions is very capable.
Based on this thought, a multilayer perceptron was added before original convolutional layer, the information between each passage is integrated, to carry
The generalization ability of model high.In practice, multilayer perceptron can be realized with 1 × 1 convolution kernel.In addition, this model is last
Sigmoid functions are not used using Maxout activation primitives in the full articulamentum of two-layer, effect of the reason with multilayer perceptron
It is similar.The expression formula of Maxout functions is Maxout (x)=max (WTx+b).What Maxout functions embodied is the think of of function approximation
Think, remove Nonlinear Function Approximation with continuous many piecewise-linear functions, burst is more, and Approximation effect is better, to nonlinear characteristic
Learning ability is stronger.This at two measure be provided to further improve the generalization ability of model, enhancing model is to nonlinear characteristic
Learning ability.The algorithm first step extracts a facial image feature for higher-dimension.Hereafter problem turns into a metric learning
Problem, is differentiated using the method for discrimination based on distance to feature, usually using Euclidean distance.It is same for belonging in training set
The image of classification (i.e. same person), it is desirable to which the Euclidean distance between similar training set is the smaller the better;Otherwise, it is desirable to it is different classes of
Training set between Euclidean distance be the bigger the better.Based on this idea, the target that can define a cost function is exactly to allow
Convolutional neural networks go to learn this cost function, so as to improve the generalization ability of model on the whole.
Given input picture x, belongs to the image x of same class (i.e. same person, hereinafter referred to as positive class) with x in training setp, instruction
It is not the image x of same class (not being same person, hereinafter referred to as bear class) that white silk is concentrated with xn, image x is by convolution for f (x) expressions
The feature that neutral net is extracted.Two threshold values are first found out on training set, defining optimization object function is:
Constraints is:
||f(x)-f(xp) | |+α < | | f (x)-f (xn)||
Wherein α is the largest interval between the positive class and negative class of f (x).This optimization problem purpose is to solve for two threshold value a,
B, if optimal solution isOrderWhen | | f (x)-f (xp) | | > a
And | | f (x)-f (xn) | | during > b, it is possible to determine that x and xpIt is same person, is not otherwise same person.Solve above mentioned problem excellent
Be can obtain after changeThen cost function can just be defined.Target be Euclidean distance between generic training set most
Small, the Euclidean distance between different classes of training set is maximum.Then problem can be described as:
Two problems are combined, is obtained:
So far, this cost function has been obtained.Problem is thus converted into one without constrained convex optimal problem, it is this kind of to ask
Topic can be solved directly using stochastic gradient descent method or quasi-Newton method.Eventually pass stochastic gradient descent method or intend newton
The weight parameter of the convolutional neural networks that method is tried to achieve is exactly optimal solution.
Extract and reform into a simple Machine Learning Problems after feature.Face verification is carried out using SVM models.
For two classification problem, SVM models have very good performance.The main policies of SVM methods are margin maximizations.From logical
For Chang Yiyi is upper, when two set to the input space are classified, it is always desirable to find one apart from the two set
All distant decision hyperplane is distinguished, because distance for distance separation hyperplane can be predicted with presentation class
Firmly believe degree, more remote apart from hyperplane, the categorised decision made is more accurate.Based on this thought, can be to from volume
The characteristic vector extracted in product neutral net goes to train a SVM model as training set.Two photos of input, through pulleying
Product neutral net is extracted and obtains two characteristic patterns, and the two characteristic patterns are input into SVM models, be may determine that when model output+1
Two pictures belong to same person, then represent that two pictures are not belonging to same person during output -1.
It is further to note that the door lock of gate control system can use electric control lock in the application, specifically gate inhibition is
The electric control lock commonly used in system includes electric mortise lock, magnetic key operated lock, electric lock mouthful etc..Wherein, electric mortise lock is mainly by two portions of lock body and lockhole
It is grouped into, the critical component of lock body is " dead bolt ".The break-make that this electric lock is exactly based on electric current drives the flexible of " dead bolt ", while
The function of coordinating " magnetic sheet " to realize locking a door or opening the door.Also the retractable of " exactly because dead bolt ", is referred to as " electric mortise lock ".
Additionally, the characteristic of its " concealed type " is suitable for the place higher to lock body confidentiality requirement.Electromagnetic lock is a kind of by electricity
Produce suction to carry out the electric control lock of door close between magnet and iron block, be powered off opening door.Common model is 280 kgfs,
Because suction is limited, may be exerted oneself to open by the big people of many people or strength.Therefore electromagnetic lock is generally used for intra-office etc.
The occasion of non-high safety rank.If for the security applications such as prison, the electromagnetic lock that more than 500 kilograms of need stretching resistance customized.
Therefore, the application can use different types of electromagnetic lock according to different application occasions.And eliminate gate control system and open door
Lock, it is the closed mode for keeping door lock that corresponding refusal eliminates gate control system.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, obtains IMAQ terminal
Collection with gate inhibition eliminate the corresponding image information of request after, can also include:
Image information is identified using deep learning photo array algorithm, if identifying that image information is to true
Face carry out shooting what is obtained, then perform carries out the step of recognition of face using deep learning face recognition algorithms to image information
Suddenly, if identifying that image information is that the face of photo is carried out shooting what is obtained, refuse to carry out face knowledge to image information
Not.
It should be noted that can carry out preventing photo or video flowing maliciously carrying out image information before processing in the application
Deception, confirms the photo content for shooting the step of be true man rather than human face photo.Specifically, there are some illegal in actual life
Molecule goes to attack face identification system using the photo or video of validated user, for this problem, some conventional solutions
Method is that system is made certain face action (such as blink, smile etc.) and recognized by voice message visitor, prevents
The photo that lawless person usurps validated user carrys out malicious attack.But, this method still has serious potential safety hazard:Illegal point
Son may also can carry out attacking system using face's high definition video steaming of validated user, and these similar approach need to increase extra in addition
Hardware device, increase the cost of system, and need user to make certain appearance body and coordinate, substantially reduce the service efficiency of user
And experience sense.Based on these considerations, there is provided a kind of to the non-intrusion type real-time judge true man of single source of photos and photo
Method.
From from the viewpoint of machine learning, this problem is simplest classification problem --- two classification, that is, judge one
The content for opening human face photo is true man or photo.If x is input picture, y is judged result --- assuming that y=1 represents input
Image is real human face, and y=0 represents that input picture is human face photo.From the angle analysis of optical imagery, real human face is that have
Three-dimensional structure, and human face photo only has two-dimensional structure;Human face photo has lacked one-dimension information relative to real human face, and its is anti-
Penetrating light should be than more uniform, and the reflected light of real human face has randomness, belongs to diffusing reflection;Both have what is differed at imaging
Depth information.Using this mechanism, using the superpower learning ability of depth convolutional network, the profound feature of image is extracted,
Two kinds of images can just be classified.
The framework of deep learning photo array algorithm disclosed in the present application can be with as shown in figure 4, wherein, convolutional neural networks
Structure input layer, convolutional layer, pond layer, convolutional layer, pond layer, full articulamentum, Logistic return layer.Wherein first volume
The convolution kernel size of lamination is 5 × 5, and port number is 6;Second convolution kernel size of convolutional layer is 5 × 5, and port number is 12;Two
Individual pond layer window size is all 2 × 2.In this network architecture using to all activated function be all sigmoid functions:
Sigmoid (z)=1/ (1+e-z)
It is h to define this network hypothesis function to be learntW, bX (), this function has special Probabilistic, it is represented
The probability that output result is equal to 1, therefore the output result of input picture x is that 1 and 0 probability is respectively:
P (y=1 | x;W, b)=hW, b(x)
P (y=0 | x;W, b)=1-hW, b(x)
Two formulas an equation can be merged into more than:
P(y|x;W, b)=(hW, b(x))y(1-hW, b(x))1-yY=0,1
Maximum-likelihood estimation is used to this equation, you can obtain loss function:
Wherein Section 2 is regularization term, it is therefore an objective to reduces the amplitude of weight, prevents overfitting.In the training process,
Convolutional neural networks first pass through propagated forward and calculate error according to above formula, then calculate partial derivative by error back propagation, from
And gradient descent method adjusting parameter can be used.Parameter during final algorithmic statement is optimal optimal models.Algorithm above energy
Single static source of photos is directly judged, compared to the method for some dynamic target trackings, judged result is more reliable, with more
Strong real-time.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, can also include:
If face recognition result does not correspond to default face or is that the face of photo is carried out shooting what is obtained, send
Carry the warning information of face recognition result or image information to designated terminal.
Sent to designated terminal by by warning information, corresponding behaviour can be carried out according to the information for obtaining by designated terminal
Make, as designated terminal correspondence user determine do not allow any other people eliminate gate control system enter in if can by specify eventually
Hold the door lock of access control system to remain off, or make if the visiting personnel of image information correspondence are designated terminal correspondence
User allows to eliminate the personnel of gate inhibition's introduction, then can by door-lock opening of designated terminal access control system etc., so as to
Enough so that the user of designated terminal realizes the remote monitoring to gate control system, and then conveniently realize corresponding control.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, transmission carries face knowledge
After warning information to the designated terminal of other result or image information, can also include:
Obtain designated terminal and receive the command information returned after warning information, perform command information and by command information and
Corresponding face recognition result or image information are stored, to detect the face recognition result or image information of storage again
The corresponding command informations of Shi Zhihang.
Wherein command information can include eliminating gate control system, keep the door lock closed mode or automatic poking of gate control system
Numbers 110 etc., can specifically be set according to actual needs, within protection scope of the present invention.Thus, by corresponding letter
After the command information that breath and designated terminal user reply is stored, can be in the later stage directly according to the command information reality of storage
Now to the treatment of correspondence image information, the control of gate control system is efficiently realized.In addition, for any people in default site
The identification events of face equipment, alert event, and other all events can be sent to back-stage management center and be shown in real time
Show, and make log recording, just in case can be with testifying when there are crime dramas, and linkage warning system pushes warning message
To APP ends, more three-dimensional security protection is realized.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, can also include:
If it is that the face of photo is shot that face recognition result does not correspond to default face or image information
, then outwardly show the information of authentication failed.
The display module for carrying out above-mentioned display can know its authentication result for visiting personnel, in addition to it is convenient with
The interaction of user, the display module can use the LCD with touch-screen, such that it is able to make the operation of this works simpler, be easy to
The use of user, user can also be by LCD come Configuration network pattern, state of setting door lock etc..
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, can also include:
If gate inhibition eliminates request to be sent by designated terminal, it indicates that gate control system eliminates gate inhibition.
The designated terminal of the limit that is possessed of control power to gate control system can be preset, therefore is asked when determining that gate inhibition eliminates
When being sent by designated terminal, can directly indicate gate control system to eliminate gate inhibition, with ensure gate control system security simultaneously, it is convenient
The use of user.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, can also include:
Judge whether that someone enters in designated area using human body infrared inductor, if it is, indicating IMAQ end
The collection of image information is held into normal mode of operation and carries out, if it is not, then indicating IMAQ terminal to keep presetting
Acquiescence park mode.
Wherein designated area can be set according to actual needs, such as within 3 meters of the door lock of gate control system.By
In the necessary non-stop run in 24 hours of whole gate control system, therefore for the consideration of power consumption, two kinds can be set for gate control system
Mode of operation:Normal mode of operation and acquiescence park mode.During normal mode of operation, all modules of whole system are all in upper
Electric working condition, power consumption is larger, and during park mode, only starts human body infrared inductor.When someone is close to equipment, human body is red
Outer induction module can sense that someone approaches, and sending signal request processor enters normal operating conditions, start all modules.
Gate control system acquiescence park mode wherein under general state, even if also can be automatic after respective operations are completed after needing normal work
Into park mode, i.e., in unmanned entrance in predeterminable area, keep park mode.
Specifically human body infrared inductor is full-automatic sensing, when people then exports high level into its induction range,
People leaves induction range, and then automatic time delay closes high level, exports low level, and system is just carried out accordingly after receiving this low level
Wake operation.The use of the module is to reduce unnecessary resource consumption.If nobody when, in Linux scheme
Process and the process not dormancy of network service as collection, then can constantly gather useless view data, be sent at backstage
Reason.The resource on backstage is so not only occupied, also increases the power consumption of this works.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, obtains IMAQ terminal
Collection with gate inhibition eliminate the corresponding image information of request after, can also include:
The ccd image information and Infrared Image Information that will be included in image information are merged, and execution utilizes deep learning
The step of face recognition algorithms carry out recognition of face to image information.
It should be noted that ccd image and infrared image respectively have its advantage and disadvantage, efficient image is become apparent from order to obtain
Both image informations are carried out image co-registration by information.Specifically, it is visible ray to use in daily life most common
Image.For human eye, it is seen that light image has abundant details and sharp color sensation, but it is in harsh climate condition
Under, the penetration capacity to air is poor, and the imaging capability at night is also poor;And infrared light contrast, it is having cigarette
Under the environmental condition of mist, penetration capacity is quite strong, at night, due to there is the temperature difference between different objects, therefore formed by it
Image remains to show the profile of object, but its shortcoming is exactly the relatively low resolution ratio being imaged.If with reference to the advantage of both photoimagings,
These multispectral information are suitably merged, then can be eliminated the image blur that environmental factor causes, and then can obtain clear
It is clear to spend enhanced target image, improve the detection to target image and recognition capability.
It is the method for being based on transform domain that current use must compare many infrared and visible ray blending algorithms, such as wavelet transformation,
Pyramid transform, contourlet transformation etc..But the above method does not possess translation invariance, image edge detailss mould is easily caused
Paste.Also a kind of non-downsampling Contourlet conversion (NCST) for possessing translation invariance, but algorithm complex is too high.By
The feature of noise and original image is all difficult to differentiate between in existing most of algorithms, it is false so as to cause the image after fusion to produce
Or fuzzy message.The present invention can be carried significantly using a kind of Image Fusion that wave conversion (NSST) is sheared based on non-lower sampling
The efficiency of algorithm high.This algorithm generates saliency map from infrared image first, then instructs infrared image according to saliency map
Target Segmentation is carried out, so background complexity or the low infrared image of signal to noise ratio can accurately be split.Then to infrared and visible
Light image carries out NSST conversion respectively, and the target area (i.e. human face region) of two images and background area are melted using different
Close strategy.The broad flow diagram of this algorithm can be as shown in Figure 5.
Wherein the infrared target region detection based on saliency map, is related to conspicuousness target detection.Infrared imaging and object
Temperature is related, therefore target area (i.e. face) is significant with respect to background area.Used here as the notable area based on frequency domain
Domain extracting method, selection Gaussian band-pass filter carrys out the notable feature of abstract image.Gaussian band-pass filter is defined as follows:
σ1,σ2(σ1>σ2) be Gaussian filter standard deviation, low-frequency cut-off frequency is by σ1Determine, high-frequency cut-off frequency is by σ2
Determine.Select suitable σ1,σ2Value, just obtains the saliency map that can keep expecting spatial frequency features.Saliency map can be under
Formula is obtained:
S (x, y)=| | Iμ- Iwhc (x, y) | |
IμIt is infrared image mean vector, Iwhc(x, y) is through the corresponding pixel value after gaussian filtering.Obtain significance
After figure, marking area that can be in saliency map selects suitable seeds pixel, carries out image segmentation.
Target area fusion rule:
In order to retain the thermal target information of infrared image as far as possible, scheme the low frequency sub-band coefficient of infrared image as fusion
The low-frequency band coefficient of picture:
LF(x, y)=Li(x, y), (x, y) ∈ T
In order to strengthen marginal information, high-frequency sub-band coefficient selection " modulus maximum ".
Wherein high-frequency sub-band and low frequency sub-band are all obtained by NSST conversion.LF,It is respectively the low frequency after fusion
Subband and high-frequency sub-band coefficient.
Background area fusion rule:Using the fusion rule based on multiresolution singular value decomposition, matrix R is carried out unusual
Value is decomposed:
R=USVT
R premultiplications UT, obtain A=UTR=SVT。
Wherein S is positive semi-definite diagonal singular value matrix.Singular value by big to minispread, larger singular value corresponding A
Above several rows, the low-frequency information in correspondence image, can largely representative image original appearance, it is several behind smaller singular value corresponding A
OK, correspondence high-frequency information, can reflect image detail.Former row elements rearrangement to A afterwards obtains low frequency sub-band, to low frequency sub-band
Continuous repetitive assignment step, just can realize multiresolution singular value decomposition.
A kind of recognition of face guard method based on deep learning provided in an embodiment of the present invention, using deep learning face
Recognizer carries out recognition of face to image information, can include:
Recognition of face is carried out to image information using the deep learning face recognition algorithms realized based on GPU.
The present invention can be divided into three parts from framework:Embedded device, backstage recognition of face server, mobile client
End.Using the processor exploitation based on ARM cortex-A series wherein on embedded device, and carry (SuSE) Linux OS.
Because the chip multimedia processing capability of cortex-A series is good, its high data throughput and high performance combination can be well
Meet network processes application, Linux is to support the operating system of multi-user, multitask, support multithreading and multi -CPU and with strong
Big network performance.(actually deep learning algorithm meter is not run directly on embedded device due to deep learning algorithm
Calculation amount is very huge, and running directly in can substantially reduce efficiency on embedded device), therefore the major function of embedded device is
Drive CCD and infrared photography head module collection image;And background server is the place for really running deep learning algorithm, with this
Meanwhile, backstage is run using GPU accelerating algorithms, improves calculating speed.Recognition result can continue through network communication protocol and return
Back to embedded device end and mobile client.Other mobile client can refer to the designated terminal in the application.
It is further to note that the data transfer in the application between different terminals can be using the mould of SDIO-WIFI
Block realizes that the network interface card meets IEEE 802.11b/g standards, it can be ensured that network data is stablized and efficiently transmitted, and its data is passed
Defeated rate is up to 54Mbps.The module uses the interface of SDIO, and the WIFI module than SPI interface is many soon.Specifically, SDIO
Bus is similar with usb bus, and SDIO buses also have two ends, and wherein one end is main frame (HOST) end, and the other end is equipment end
(DEVICE) it is, that, in order to simplify the design of DEVICE, all of communication is all by HOST using design as HOST-DEVICE
End sends what order started.As long as the order of HOST can be parsed at DEVICE ends, it is possible to carry out communicating with HOST, SDIO's
HOST can connect multiple DEVICE.Because system operation needs frequently to send network request, and server background enters line number
According to interaction, for network congestion causes program to enter unlimited wait state in preventing network transmission process, system uses multithreading journey
The thought of sequence design, thread object is all opened up when network request is carried out every time temporarily, and passes through message machine after being returned
System notifies main thread, allows main thread to parse returning result and visual feedback is to user, and releasing network request thread resource.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, as shown in fig. 6, can be with
Including:
First judge module 11, is used for:Receive gate inhibition and eliminate request, judge that gate inhibition eliminates whether request is sent out by designated terminal
Send, if it is not, then determining that gate inhibition eliminates request and sent by IMAQ terminal;
Image processing module 12, is used for:The image corresponding with gate inhibition's elimination request for obtaining the collection of IMAQ terminal is believed
Breath, recognition of face is carried out using deep learning face recognition algorithms to image information, obtains corresponding face recognition result;
Second judge module 13, is used for:Whether the corresponding default face of face recognition result is judged, if it is, indicating door
Access control system eliminates gate inhibition, if it is not, then refusal eliminates gate inhibition.
A kind of face recognition door control system based on deep learning provided in an embodiment of the present invention, can also include:
3rd judge module, is used for:Obtain the collection of IMAQ terminal eliminates the corresponding image information of request with gate inhibition
Afterwards, image information is identified using deep learning photo array algorithm, if identifying that image information is to real
Face carries out shooting what is obtained, then perform carries out the step of recognition of face using deep learning face recognition algorithms to image information
Suddenly, if identifying that image information is that the face of photo is carried out shooting what is obtained, refuse to carry out face knowledge to image information
Not.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, can also include:
Alarm modules, are used for:If face recognition result does not correspond to the face of default face or image information to photo
Carry out shooting what is obtained, then send and carry the warning information of face recognition result or image information to designated terminal.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, can also include:
Memory module, is used for:Transmission carry the warning information of face recognition result or image information to designated terminal it
Afterwards, obtain designated terminal and receive the command information returned after warning information, perform command information and by command information and correspondence
Face recognition result or image information stored, held with when the face recognition result or image information of storage is detected again
The corresponding command information of row.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, can also include:
Display module, is used for:If face recognition result does not correspond to the face of default face or image information to photo
Carry out shooting what is obtained, then outwardly show the information of authentication failed.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, can also include:
Gate inhibition's cancellation module, is used for:If gate inhibition eliminates request to be sent by designated terminal, it indicates that gate control system disappears
Except gate inhibition.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, can also include:
4th judge module, is used for:Judge whether that someone enters in designated area using human body infrared inductor, if
It is, it indicates that IMAQ terminal is into normal mode of operation and carries out the collection of image information, if it is not, then indicating image to adopt
Collection terminal keeps acquiescence park mode set in advance.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, can also include:
Fusion Module, is used for:After what acquisition IMAQ terminal was gathered eliminates the corresponding image information of request with gate inhibition,
The ccd image information and Infrared Image Information that will be included in image information are merged, and are performed and are utilized deep learning recognition of face
The step of algorithm carries out recognition of face to image information.
The embodiment of the present invention additionally provides a kind of face recognition door control system based on deep learning, image processing module bag
Include:
Graphics processing unit, is used for:Image information is carried out using the deep learning face recognition algorithms realized based on GPU
Recognition of face.
The embodiment of the present invention additionally provides a kind of saying for relevant portion in face recognition door control system based on deep learning
The bright embodiment of the present invention that refers to additionally provides a kind of the detailed of corresponding part in recognition of face guard method based on deep learning
Describe in detail bright, will not be repeated here.
The foregoing description of the disclosed embodiments, enables those skilled in the art to realize or uses the present invention.To this
Various modifications of a little embodiments will be apparent for a person skilled in the art, and generic principles defined herein can
Without departing from the spirit or scope of the present invention, to realize in other embodiments.Therefore, the present invention will not be limited
It is formed on the embodiments shown herein, and is to fit to consistent with principles disclosed herein and features of novelty most wide
Scope.
Claims (10)
1. a kind of recognition of face guard method based on deep learning, it is characterised in that including:
Receive gate inhibition and eliminate request, judge that the gate inhibition eliminates whether request is sent by designated terminal, if it is not, then determining described
Gate inhibition eliminates request and is sent by IMAQ terminal;
Obtain the collection of described image acquisition terminal eliminates the corresponding image information of request with the gate inhibition, using deep learning people
Face recognizer carries out recognition of face to described image information, obtains corresponding face recognition result;
Judge the whether corresponding default face of the face recognition result, if it is, indicate gate control system to eliminate gate inhibition, if
No, then refusal eliminates gate inhibition.
2. method according to claim 1, it is characterised in that obtaining the collection of described image acquisition terminal with the gate inhibition
Eliminate after the corresponding image information of request, also include:
Described image information is identified using deep learning photo array algorithm, if identifying that described image information is right
Real face is carried out shooting what is obtained, then perform described is carried out using deep learning face recognition algorithms to described image information
The step of recognition of face, if identifying that described image information is that the face of photo is carried out shooting what is obtained, refuse to institute
Stating image information carries out recognition of face.
3. method according to claim 2, it is characterised in that also include:
If it is that the face of photo is shot that the face recognition result does not correspond to default face or described image information
Obtain, then send and carry the warning information of the face recognition result or described image information to the designated terminal.
4. method according to claim 3, it is characterised in that transmission carries the face recognition result or described image
After the warning information of information to the designated terminal, also include:
Obtain the designated terminal and receive the command information returned after the warning information, perform the command information and by institute
State command information and corresponding face recognition result or image information is stored, know with the face for detecting storage again
Corresponding command information is performed when other result or described image information.
5. method according to claim 2, it is characterised in that also include:
If it is that the face of photo is shot that the face recognition result does not correspond to default face or described image information
Obtain, then outwardly show the information of authentication failed.
6. method according to claim 2, it is characterised in that also include:
If the gate inhibition eliminates request and is sent by the designated terminal, it indicates that the gate control system eliminates gate inhibition.
7. method according to claim 1, it is characterised in that also include:
Judge whether that someone enters in designated area using human body infrared inductor, if it is, indicating described image collection eventually
The collection of image information is held into normal mode of operation and carries out, if it is not, then indicating described image acquisition terminal to keep advance
The acquiescence park mode of setting.
8. method according to claim 1, it is characterised in that obtaining the collection of described image acquisition terminal with the gate inhibition
Eliminate after the corresponding image information of request, also include:
The ccd image information and Infrared Image Information that will be included in described image information are merged, and perform the utilization depth
The step of study face recognition algorithms carry out recognition of face to described image information.
9. method according to claim 8, it is characterised in that believed described image using deep learning face recognition algorithms
Breath carries out recognition of face, including:
Recognition of face is carried out to described image information using the deep learning face recognition algorithms realized based on GPU.
10. a kind of face recognition door control system based on deep learning, it is characterised in that including:
First judge module, is used for:Receive gate inhibition and eliminate request, judge that the gate inhibition eliminates whether request is sent out by designated terminal
Send, if it is not, then determining that the gate inhibition eliminates request and sent by IMAQ terminal;
Image processing module, is used for:Obtain the collection of described image acquisition terminal eliminates the corresponding image of request with the gate inhibition
Information, recognition of face is carried out using deep learning face recognition algorithms to described image information, obtains corresponding recognition of face knot
Really;
Second judge module, is used for:Whether the corresponding default face of the face recognition result is judged, if it is, indicating gate inhibition
System eliminates gate inhibition, if it is not, then refusal eliminates gate inhibition.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710192816.2A CN106803301A (en) | 2017-03-28 | 2017-03-28 | A kind of recognition of face guard method and system based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710192816.2A CN106803301A (en) | 2017-03-28 | 2017-03-28 | A kind of recognition of face guard method and system based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106803301A true CN106803301A (en) | 2017-06-06 |
Family
ID=58981762
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710192816.2A Pending CN106803301A (en) | 2017-03-28 | 2017-03-28 | A kind of recognition of face guard method and system based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106803301A (en) |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107366491A (en) * | 2017-07-18 | 2017-11-21 | 广东工业大学 | The control method and electric extendable door of a kind of electric extendable door |
CN107437081A (en) * | 2017-08-07 | 2017-12-05 | 北京中星微电子有限公司 | Face identification method, device and storage medium based on depth volume neutral net |
CN108182746A (en) * | 2018-01-30 | 2018-06-19 | 百度在线网络技术(北京)有限公司 | Control system, method and apparatus |
CN108287162A (en) * | 2018-01-09 | 2018-07-17 | 温州三特食品科技有限公司 | A kind of method of food security intelligent measurement |
CN108364374A (en) * | 2017-12-28 | 2018-08-03 | 武汉烽火众智数字技术有限责任公司 | Face access control device based on deep learning and method |
CN108492411A (en) * | 2018-03-13 | 2018-09-04 | 东莞市友联奕诺生物科技有限公司 | A kind of electronic lock control system |
CN108932492A (en) * | 2018-06-28 | 2018-12-04 | 福州昌宇五金锁具制品有限公司 | A kind of image fingerprint extracting method based on non-sampled shearing wave conversion |
CN109033988A (en) * | 2018-06-29 | 2018-12-18 | 江苏食品药品职业技术学院 | A kind of library's access management system based on recognition of face |
CN109243021A (en) * | 2018-08-28 | 2019-01-18 | 余利 | Deeply learning type intelligent door lock system and device based on user experience analysis |
CN109377628A (en) * | 2018-12-17 | 2019-02-22 | 深圳市恩钛控股有限公司 | A kind of intelligent access control system and method |
CN109461231A (en) * | 2018-10-15 | 2019-03-12 | 珠海格力电器股份有限公司 | A kind of control method for door lock, device, control equipment and readable storage medium storing program for executing |
CN109977974A (en) * | 2017-12-28 | 2019-07-05 | 同方威视科技江苏有限公司 | The online experience platform of image detection and the image detecting method for using it |
TWI684923B (en) * | 2016-07-22 | 2020-02-11 | 美商美國Nec實驗室有限公司 | Liveness detection for antispoof face recognition |
CN111027397A (en) * | 2019-11-14 | 2020-04-17 | 上海交通大学 | Method, system, medium and device for detecting comprehensive characteristic target in intelligent monitoring network |
CN111583481A (en) * | 2020-04-21 | 2020-08-25 | 中国商用飞机有限责任公司 | Access control system for cockpit door of aircraft and aircraft |
CN111798594A (en) * | 2020-06-30 | 2020-10-20 | 德施曼机电(中国)有限公司 | Intelligent door lock and unlocking control method and storage medium thereof |
CN112131997A (en) * | 2020-09-17 | 2020-12-25 | 南京智凝人工智能研究院有限公司 | Face recognition system and method based on deep learning |
CN112818901A (en) * | 2021-02-22 | 2021-05-18 | 成都睿码科技有限责任公司 | Wearing mask face recognition method based on eye attention mechanism |
CN113066214A (en) * | 2021-03-26 | 2021-07-02 | 深圳市博盛科电子有限公司 | Access control system based on 5G network remote monitoring |
CN113192240A (en) * | 2021-03-12 | 2021-07-30 | 广州朗国电子科技有限公司 | Deep learning-based identification module identification method, equipment and medium |
CN113947554A (en) * | 2020-07-17 | 2022-01-18 | 四川大学 | Multi-focus image fusion method based on NSST and significant information extraction |
CN114202831A (en) * | 2021-12-08 | 2022-03-18 | 广东好太太智能家居有限公司 | Household care alarm method, system, equipment and medium based on intelligent lock |
CN115130082A (en) * | 2022-08-26 | 2022-09-30 | 中电长城圣非凡信息系统有限公司 | Intelligent sensing and safety control method for ruggedized computer |
CN117058803A (en) * | 2023-10-13 | 2023-11-14 | 浪潮智慧科技创新(山东)有限公司 | Intelligent data acquisition method and system based on deep learning |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103024281A (en) * | 2013-01-11 | 2013-04-03 | 重庆大学 | Infrared and visible video integration system |
CN103414814A (en) * | 2013-08-16 | 2013-11-27 | 北京小米科技有限责任公司 | Picture processing method and device and terminal device |
CN105912908A (en) * | 2016-04-14 | 2016-08-31 | 苏州优化智能科技有限公司 | Infrared-based real person living body identity verification method |
CN106204815A (en) * | 2016-06-23 | 2016-12-07 | 江西洪都航空工业集团有限责任公司 | A kind of gate control system based on human face detection and recognition |
CN106296950A (en) * | 2016-09-30 | 2017-01-04 | 深圳市商汤科技有限公司 | A kind of gate control system based on recognition of face |
-
2017
- 2017-03-28 CN CN201710192816.2A patent/CN106803301A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103024281A (en) * | 2013-01-11 | 2013-04-03 | 重庆大学 | Infrared and visible video integration system |
CN103414814A (en) * | 2013-08-16 | 2013-11-27 | 北京小米科技有限责任公司 | Picture processing method and device and terminal device |
CN105912908A (en) * | 2016-04-14 | 2016-08-31 | 苏州优化智能科技有限公司 | Infrared-based real person living body identity verification method |
CN106204815A (en) * | 2016-06-23 | 2016-12-07 | 江西洪都航空工业集团有限责任公司 | A kind of gate control system based on human face detection and recognition |
CN106296950A (en) * | 2016-09-30 | 2017-01-04 | 深圳市商汤科技有限公司 | A kind of gate control system based on recognition of face |
Non-Patent Citations (5)
Title |
---|
K.SHAILAJA: "Effective Face Recognition using Deep Learning based Linear Discriminant Classification", 《2016 IEEE INTERNATIONAL CONFERENCE ON COMPUTATIONAL INTELLIGENCE AND COMPUTING RESEARCH (ICCIC)》, 30 June 2016 (2016-06-30) * |
张家良: "红外与可见光图像配准及融合技术研究", 《中国优秀硕士学位论文全文数据库》, 30 June 2013 (2013-06-30), pages 47 - 49 * |
申小敏: "基于卷积神经网络的大规模人脸聚类", 《中国优秀硕士学位论文全文数据库》, 15 November 2016 (2016-11-15), pages 51 - 55 * |
胡亚君: "卷积神经网络在人脸识别的应用", 《无线互联科技》, 10 May 2016 (2016-05-10) * |
蔡念: "鲁棒主成分分析的运动目标检测综述", 《中国图象图形学报》, 16 October 2016 (2016-10-16), pages 1265 - 1275 * |
Cited By (32)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI684923B (en) * | 2016-07-22 | 2020-02-11 | 美商美國Nec實驗室有限公司 | Liveness detection for antispoof face recognition |
CN107366491A (en) * | 2017-07-18 | 2017-11-21 | 广东工业大学 | The control method and electric extendable door of a kind of electric extendable door |
CN107437081A (en) * | 2017-08-07 | 2017-12-05 | 北京中星微电子有限公司 | Face identification method, device and storage medium based on depth volume neutral net |
CN108364374A (en) * | 2017-12-28 | 2018-08-03 | 武汉烽火众智数字技术有限责任公司 | Face access control device based on deep learning and method |
CN109977974A (en) * | 2017-12-28 | 2019-07-05 | 同方威视科技江苏有限公司 | The online experience platform of image detection and the image detecting method for using it |
CN108287162A (en) * | 2018-01-09 | 2018-07-17 | 温州三特食品科技有限公司 | A kind of method of food security intelligent measurement |
CN108182746A (en) * | 2018-01-30 | 2018-06-19 | 百度在线网络技术(北京)有限公司 | Control system, method and apparatus |
CN108492411A (en) * | 2018-03-13 | 2018-09-04 | 东莞市友联奕诺生物科技有限公司 | A kind of electronic lock control system |
CN108932492A (en) * | 2018-06-28 | 2018-12-04 | 福州昌宇五金锁具制品有限公司 | A kind of image fingerprint extracting method based on non-sampled shearing wave conversion |
CN109033988A (en) * | 2018-06-29 | 2018-12-18 | 江苏食品药品职业技术学院 | A kind of library's access management system based on recognition of face |
CN109243021A (en) * | 2018-08-28 | 2019-01-18 | 余利 | Deeply learning type intelligent door lock system and device based on user experience analysis |
CN109243021B (en) * | 2018-08-28 | 2021-09-17 | 余利 | Deep reinforcement learning type intelligent door lock system and device based on user experience analysis |
CN109461231A (en) * | 2018-10-15 | 2019-03-12 | 珠海格力电器股份有限公司 | A kind of control method for door lock, device, control equipment and readable storage medium storing program for executing |
CN109377628A (en) * | 2018-12-17 | 2019-02-22 | 深圳市恩钛控股有限公司 | A kind of intelligent access control system and method |
CN111027397A (en) * | 2019-11-14 | 2020-04-17 | 上海交通大学 | Method, system, medium and device for detecting comprehensive characteristic target in intelligent monitoring network |
CN111027397B (en) * | 2019-11-14 | 2023-05-12 | 上海交通大学 | Comprehensive feature target detection method, system, medium and equipment suitable for intelligent monitoring network |
CN111583481A (en) * | 2020-04-21 | 2020-08-25 | 中国商用飞机有限责任公司 | Access control system for cockpit door of aircraft and aircraft |
CN111583481B (en) * | 2020-04-21 | 2021-11-19 | 中国商用飞机有限责任公司 | Access control system for cockpit door of aircraft and aircraft |
CN111798594A (en) * | 2020-06-30 | 2020-10-20 | 德施曼机电(中国)有限公司 | Intelligent door lock and unlocking control method and storage medium thereof |
CN113947554B (en) * | 2020-07-17 | 2023-07-14 | 四川大学 | Multi-focus image fusion method based on NSST and significant information extraction |
CN113947554A (en) * | 2020-07-17 | 2022-01-18 | 四川大学 | Multi-focus image fusion method based on NSST and significant information extraction |
CN112131997A (en) * | 2020-09-17 | 2020-12-25 | 南京智凝人工智能研究院有限公司 | Face recognition system and method based on deep learning |
CN112818901B (en) * | 2021-02-22 | 2023-04-07 | 成都睿码科技有限责任公司 | Wearing mask face recognition method based on eye attention mechanism |
CN112818901A (en) * | 2021-02-22 | 2021-05-18 | 成都睿码科技有限责任公司 | Wearing mask face recognition method based on eye attention mechanism |
CN113192240A (en) * | 2021-03-12 | 2021-07-30 | 广州朗国电子科技有限公司 | Deep learning-based identification module identification method, equipment and medium |
CN113066214A (en) * | 2021-03-26 | 2021-07-02 | 深圳市博盛科电子有限公司 | Access control system based on 5G network remote monitoring |
CN114202831A (en) * | 2021-12-08 | 2022-03-18 | 广东好太太智能家居有限公司 | Household care alarm method, system, equipment and medium based on intelligent lock |
CN114202831B (en) * | 2021-12-08 | 2024-02-09 | 广东好太太智能家居有限公司 | Household care alarm method, system, equipment and medium based on intelligent lock |
CN115130082A (en) * | 2022-08-26 | 2022-09-30 | 中电长城圣非凡信息系统有限公司 | Intelligent sensing and safety control method for ruggedized computer |
CN115130082B (en) * | 2022-08-26 | 2022-11-04 | 中电长城圣非凡信息系统有限公司 | Intelligent sensing and safety control method for ruggedized computer |
CN117058803A (en) * | 2023-10-13 | 2023-11-14 | 浪潮智慧科技创新(山东)有限公司 | Intelligent data acquisition method and system based on deep learning |
CN117058803B (en) * | 2023-10-13 | 2024-01-05 | 浪潮智慧科技创新(山东)有限公司 | Intelligent data acquisition method and system based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106803301A (en) | A kind of recognition of face guard method and system based on deep learning | |
CN108182409B (en) | Living body detection method, living body detection device, living body detection equipment and storage medium | |
CN106372576A (en) | Deep learning-based intelligent indoor intrusion detection method and system | |
US6661907B2 (en) | Face detection in digital images | |
CN107230267B (en) | Intelligence In Baogang Kindergarten based on face recognition algorithms is registered method | |
CN106845328B (en) | A kind of Intelligent human-face recognition methods and system based on dual camera | |
CN110472519B (en) | Human face in-vivo detection method based on multiple models | |
CN102844766A (en) | Human eyes images based multi-feature fusion identification method | |
CN102902959A (en) | Face recognition method and system for storing identification photo based on second-generation identity card | |
CN109190475A (en) | A kind of recognition of face network and pedestrian identify network cooperating training method again | |
CN109145742A (en) | A kind of pedestrian recognition method and system | |
CN101980242A (en) | Human face discrimination method and system and public safety system | |
CN101739742A (en) | Networking type multi-channel access control and attendance system | |
CN110276320A (en) | Guard method, device, equipment and storage medium based on recognition of face | |
CN108268850A (en) | A kind of big data processing method based on image | |
CN107609515B (en) | Double-verification face comparison system and method based on Feiteng platform | |
WO2022121498A1 (en) | Identity recognition method, model training method, apparatuses, and device and storage medium | |
CN108446690A (en) | A kind of human face in-vivo detection method based on various visual angles behavioral characteristics | |
CN107516076A (en) | Portrait identification method and device | |
CN105184236A (en) | Robot-based face identification system | |
CN106886771A (en) | The main information extracting method of image and face identification method based on modularization PCA | |
CN103942545A (en) | Method and device for identifying faces based on bidirectional compressed data space dimension reduction | |
Lee et al. | Distinction Between Real Faces and Photos by Analysis of Face Data. | |
WO2023093241A1 (en) | Pedestrian re-identification method and apparatus, and storage medium | |
CN116805436A (en) | Intelligent access control, control system and control 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170606 |