CN106650658A - Method and device for identifying user identity - Google Patents
Method and device for identifying user identity Download PDFInfo
- Publication number
- CN106650658A CN106650658A CN201611169854.8A CN201611169854A CN106650658A CN 106650658 A CN106650658 A CN 106650658A CN 201611169854 A CN201611169854 A CN 201611169854A CN 106650658 A CN106650658 A CN 106650658A
- Authority
- CN
- China
- Prior art keywords
- depth characteristic
- characteristic vector
- human face
- depth
- prestores
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Abstract
The invention provides a method and a device for identifying a user identity. The method comprises the following steps of: detecting a face picture; inputting the face picture into a face identification model, and carrying out calculation by the face identification model to obtain the depth feature vector of the face picture; calculating a distance value independently between the depth feature vector and each pre-stored depth feature vector in a pre-stored depth feature vector group, wherein the pre-stored depth feature vector group comprises a plurality of pre-stored depth feature vectors, and the pre-stored depth feature vectors correspond to user identity information one by one; determining a corresponding target pre-stored depth feature vector when the distance value is minimum; and determining the user identity information corresponding to the target pre-stored depth feature vector as the user identity in the face picture. By use of the method, a recognition rate and accuracy for a robot to identify the user identity are improved.
Description
Technical field
The present invention relates to technical field of face recognition, more particularly to a kind of robot identifying user body based on deep learning
The method and apparatus of part.
Background technology
With the high speed development of science and technology, robot building technology also quickly grown, the application of robot by
Stepping enters home services industry.
Service for infrastructure robot be it is a kind of do not affected by environment, temperature, can dynamically realize to user identity identification
Robot, such robot needs the identification capability for having brilliance in terms of user identity identification.
Therefore, discrimination, the accuracy for how improving robot identifying user identity is very crucial.
The content of the invention
In view of this, the present invention provides a kind of method and apparatus of identifying user identity, to improve robot identifying user
The discrimination of identity, accuracy.Technical scheme is as follows：
Based on an aspect of of the present present invention, the present invention provides a kind of method of identifying user identity, including：
Detection face picture；
The face picture is input into human face recognition model, using the human face recognition model face is calculated
The depth characteristic vector of picture；
Calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each prestore depth characteristic to
The distance between amount value, the depth characteristic Vector Groups for prestoring include multiple depth characteristic vectors that prestore, the depth that prestores
Characteristic vector is corresponded with subscriber identity information；
Corresponding target prestores depth characteristic vector when determining that distance value is minimum；
Determine that the target depth characteristic corresponding subscriber identity information of vector that prestores is user in the face picture
Identity.
Preferably, the human face recognition model is depth convolutional neural networks Deep CNN human face recognition models.
Preferably, before the detection face picture, methods described also includes：
Based on substantial amounts of human face data, the substantial amounts of face number is trained using propagated forward and the method for back-propagating
According to obtain the Deep CNN human face recognition models.
Preferably, it is described calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each prestore
The distance between depth characteristic vector value includes：
It is vectorial with the depth characteristic for prestoring respectively using the depth characteristic vector is calculated closest to KNN minimum distances algorithm
Each in group prestores the distance between depth characteristic vector value.
Based on another aspect of the present invention, the present invention also provides a kind of device of identifying user identity, including：
Picture detector unit, for detecting face picture；
Picture processing unit, for the face picture to be input into default human face recognition model, using the face
Identification model is calculated the depth characteristic vector of the face picture；
Metrics calculation unit, for calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in it is every
Individual the distance between the depth characteristic vector value that prestores, the depth characteristic Vector Groups for prestoring include it is multiple prestore depth characteristic to
Amount, the depth characteristic vector that prestores is corresponded with subscriber identity information；
First determining unit, prestores depth characteristic vector for corresponding target when determining that distance value is minimum；
Second determining unit, for determining that the target depth characteristic corresponding subscriber identity information of vector that prestores is described
The identity of user in face picture.
Preferably, the human face recognition model is depth convolutional neural networks Deep CNN human face recognition models.
Preferably, also include：
Model training unit, for based on substantial amounts of human face data, being instructed using propagated forward and the method for back-propagating
The white silk substantial amounts of human face data, to obtain the Deep CNN human face recognition models.
Preferably, the metrics calculation unit using closest to KNN minimum distances algorithm specifically for calculating the depth
Characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each the distance between depth characteristic vector value that prestores.
The method of the identifying user identity that the present invention is provided includes：Detection face picture；By the face picture be input into
Default human face recognition model, using the human face recognition model depth characteristic vector of the face picture is calculated；Meter
Calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each prestore between depth characteristic vector away from
From value, the depth characteristic Vector Groups for prestoring include it is multiple prestore depth characteristics vector, it is described prestore depth characteristic vector with
Subscriber identity information is corresponded；Corresponding target prestores depth characteristic vector when determining that distance value is minimum；Determine the mesh
Mark the identity that the corresponding subscriber identity information of depth characteristic vector is user in the face picture that prestores.The present invention utilizes face
Identification model is calculated the depth characteristic vector of face picture, and then special with the depth that respectively prestores by calculating depth characteristic vector
The distance value between vector is levied, corresponding target prestores depth characteristic vector when determining that distance value is minimum, the target is prestored depth
The corresponding subscriber identity information of degree characteristic vector is the identity of user in the face picture, and the present invention improves robot identification
The discrimination of user identity, accuracy.
Description of the drawings
In order to be illustrated more clearly that 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 be with basis
The accompanying drawing of offer obtains other accompanying drawings.
A kind of flow chart of the method for identifying user identity that Fig. 1 is provided for the present invention；
A kind of structural representation of the device of identifying user identity that Fig. 2 is provided for the present invention.
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 the embodiment of whole.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, a kind of flow chart of the method for identifying user identity of present invention offer is provided, including：
Step 101, detects face picture.
Wherein, face picture refers to include the picture of face information.General, the present invention can take the photograph using shooting is first-class
As device catch face picture, after camera head captures face picture, the present invention can from camera head active obtaining
Or passively receive the face picture.
Step 102, the face picture is input into human face recognition model, is calculated using the human face recognition model
The depth characteristic vector of the face picture.
In the present invention, human face recognition model is specially Deep CNN (Deep Convolutional Neural
Networks, depth convolutional neural networks) human face recognition model.
Particularly, should also include before step 101 of the present invention is realized：Step 100, based on substantial amounts of human face data, adopts
Propagated forward (Forward Propagation) and the method for back-propagating (Back Propagation) are described a large amount of to train
Human face data, to obtain the Deep CNN human face recognition models.
Deep CNN human face recognition models in the present invention be based on multi-layer C NN (>10 layers) million grades of parameters of framework
Deep CNN human face recognition models, the basis of the Deep CNN human face recognition models is depth convolutional neural networks, the depth volume
Product neutral net has 37 layers, including 16 convolutional layers, and includes million grades of parameter in the network.Wrap wherein in the network
The face figure (hundreds of thousands face rank) that the parameter of million grades for including must pass through to have demarcated in a large number constantly repeatedly treats supervised training
Practise (supervised learning), adjustment model parameter value is minimized with reaching tranining database face identification rate.
In order that model parameter can better discriminate between the different faces of different geographical, such as Asia face, training sample
Concentration needs to collect substantial amounts of Asia human face data, by the training of the substantial amounts of Asia human face data to collecting, just can guarantee that
The depth characteristic vector of the bottommost layer for obtaining has the ability of the different faces of very strong differentiation.
It is concrete in embodiments of the present invention, it is necessary first to set the structure of Deep CNN human face recognition models, i.e., this
The deep learning neutral net of bright employing multilayer, including convolutional layer, pooling layers, relu layers, full articulamentum, softmax classification
And triploss layers.Be in training process training convolutional layer kernel function, full articulamentum link power, softmax classification layer and
The parameter of triplet loss layers.In actual application of the present invention, using the human face photo data after normalization as defeated
Enter data, using the nominal data of corresponding face as output, with softmax loss functions as object function.Specifically such as, by people
The testing result of face picture normalizes to 224 × 224 after calibration, is input to after Deep CNN human face recognition models and just may be used
Obtain depth characteristic vector.
Step 103, calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each depth that prestores
The distance between degree characteristic vector value.
Wherein, the depth characteristic Vector Groups for prestoring include multiple depth characteristic vectors that prestore, and the depth that prestores is special
Levy vector to correspond with subscriber identity information.
In actual application of the present invention, face one depth characteristic vector of correspondence of each user, the present invention can
So that the corresponding depth characteristic vector of each user is pre-stored in depth characteristic Vector Groups, thus the depth characteristic in the present invention to
Amount group includes the depth characteristic vector of thousands of dimensions.
Specifically, the present invention can be using KNN (k-Nearest Neighbor, closest to) minimum distance algorithm, successively
Calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each prestore between depth characteristic vector
Distance value.KNN is classical sorting algorithm:The classification of input sample is determined by " majority voting " that its nearest K neighbour classifies.
As K=1, the classification of sample is assigned to the classification of its nearest-neighbors.
In the present invention, through mass data supervised training, between the depth characteristic vector of the face picture of same people away from
Close to the distance between the face picture of (such as Euclidean distance) than different people, i.e., the depth characteristic vector of the face picture of same people
The distance between value can be minimum.Therefore the present invention is special with the depth for prestoring by the depth characteristic vector of calculated face picture
Each levied in Vector Groups prestores depth characteristic vector matching primitives distance value one by one.
Step 104, corresponding target prestores depth characteristic vector when determining that distance value is minimum.
Step 105, determines that the target depth characteristic corresponding subscriber identity information of vector that prestores is the face picture
The identity of middle user.
Therefore the method for the identifying user identity that the application present invention is provided, by using human face recognition model people is calculated
The depth characteristic vector of face picture, and then the distance value by calculating depth characteristic vector and respectively prestore between depth characteristic vector,
Corresponding target prestores depth characteristic vector when determining that distance value is minimum, and the target is prestored the corresponding use of depth characteristic vector
Family identity information is the identity of user in the face picture, the discrimination that improve robot identifying user identity of the invention,
Accuracy.
Based on a kind of method of identifying user identity that the present invention is provided above, the present invention also provides a kind of identifying user body
The device of part, as shown in Fig. 2 including：
Picture detector unit 100, for detecting face picture；
Picture processing unit 200, for the face picture to be input into default human face recognition model, using the people
Face identification model is calculated the depth characteristic vector of the face picture；
Metrics calculation unit 300, for calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in
Each the distance between depth characteristic vector value that prestores, the depth characteristic Vector Groups for prestoring include that multiple depth that prestore are special
Vector is levied, the depth characteristic vector that prestores is corresponded with subscriber identity information；
First determining unit 400, prestores depth characteristic vector for corresponding target when determining that distance value is minimum；
Second determining unit 500, for determining that the target depth characteristic corresponding subscriber identity information of vector that prestores is
The identity of user in the face picture.
Wherein, the human face recognition model is Deep CNN human face recognition models.
Described device also includes：
Model training unit 600, for based on substantial amounts of human face data, using propagated forward and the method for back-propagating come
The substantial amounts of human face data is trained, to obtain the Deep CNN human face recognition models.
The metrics calculation unit 300 using KNN minimum distances algorithm specifically for calculating the depth characteristic vector point
Not with the depth characteristic Vector Groups for prestoring in each the distance between depth characteristic vector value that prestores.
It should be noted that each embodiment in this specification is described by the way of progressive, each embodiment weight
Point explanation is all difference with other embodiment, between each embodiment identical similar part mutually referring to.
For device class embodiment, due to itself and embodiment of the method basic simlarity, so description is fairly simple, related part ginseng
See the part explanation of embodiment of the method.
Finally, in addition it is also necessary to explanation, herein, such as first and second or the like relational terms be used merely to by
One entity or operation make a distinction with another entity or operation, and not necessarily require or imply these entities or operation
Between there is any this actual relation or order.And, term " including ", "comprising" or its any other variant meaning
Covering including for nonexcludability, so that a series of process, method, article or equipment including key elements not only includes that
A little key elements, but also including other key elements being not expressly set out, or also include for this process, method, article or
The intrinsic key element of equipment.In the absence of more restrictions, the key element for being limited by sentence "including a ...", does not arrange
Except also there is other identical element in including the process of the key element, method, article or equipment.
A kind of method and apparatus of identifying user identity provided by the present invention is described in detail above, herein
Apply specific case to be set forth the principle and embodiment of the present invention, the explanation of above example is only intended to help
Understand the method for the present invention and its core concept；Simultaneously for one of ordinary skill in the art, according to the thought of the present invention,
Will change in specific embodiments and applications, in sum, this specification content should not be construed as to this
The restriction of invention.
Claims (8)
1. a kind of method of identifying user identity, it is characterised in that include：
Detection face picture；
The face picture is input into human face recognition model, using the human face recognition model face picture is calculated
Depth characteristic vector；
Calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each depth characteristic vector that prestores
Between distance value, the depth characteristic Vector Groups for prestoring include it is multiple prestore depth characteristics vector, the depth characteristic that prestores
Vector is corresponded with subscriber identity information；
Corresponding target prestores depth characteristic vector when determining that distance value is minimum；
Determine the identity for the target prestoring the depth characteristic corresponding subscriber identity information of vector being user in the face picture.
2. method according to claim 1, it is characterised in that the human face recognition model is depth convolutional neural networks
Deep CNN human face recognition models.
3. method according to claim 2, it is characterised in that before the detection face picture, methods described also includes：
Based on substantial amounts of human face data, the substantial amounts of human face data is trained using propagated forward and the method for back-propagating,
To obtain the Deep CNN human face recognition models.
4. the method according to any one of claim 1-3, it is characterised in that the calculating depth characteristic vector is respectively
Include with the distance between the depth characteristic vector value that prestores of each in the depth characteristic Vector Groups for prestoring：
Using closest to KNN minimum distances algorithm calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in
Each the distance between depth characteristic vector value that prestores.
5. a kind of device of identifying user identity, it is characterised in that include：
Picture detector unit, for detecting face picture；
Picture processing unit, for the face picture to be input into default human face recognition model, using the recognition of face
Model is calculated the depth characteristic vector of the face picture；
Metrics calculation unit, for calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each is pre-
The distance between depth characteristic vector value is deposited, the depth characteristic Vector Groups for prestoring include multiple depth characteristic vectors that prestore,
The depth characteristic vector that prestores is corresponded with subscriber identity information；
First determining unit, prestores depth characteristic vector for corresponding target when determining that distance value is minimum；
Second determining unit, for determining that the target depth characteristic corresponding subscriber identity information of vector that prestores is the face
The identity of user in picture.
6. device according to claim 5, it is characterised in that the human face recognition model is depth convolutional neural networks
Deep CNN human face recognition models.
7. device according to claim 6, it is characterised in that also include：
Model training unit, for based on substantial amounts of human face data, using propagated forward and the method for back-propagating to train
Substantial amounts of human face data is stated, to obtain the Deep CNN human face recognition models.
8. the device according to any one of claim 5-7, it is characterised in that the metrics calculation unit is specifically for adopting
With closest to KNN minimum distances algorithm calculate depth characteristic vector respectively with the depth characteristic Vector Groups for prestoring in each
The distance between the depth characteristic that prestores vector value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611169854.8A CN106650658A (en) | 2016-12-16 | 2016-12-16 | Method and device for identifying user identity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611169854.8A CN106650658A (en) | 2016-12-16 | 2016-12-16 | Method and device for identifying user identity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106650658A true CN106650658A (en) | 2017-05-10 |
Family
ID=58822164
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611169854.8A Pending CN106650658A (en) | 2016-12-16 | 2016-12-16 | Method and device for identifying user identity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650658A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108492575A (en) * | 2018-04-11 | 2018-09-04 | 济南浪潮高新科技投资发展有限公司 | A kind of intelligent vehicle type identifier method |
| CN109117801A (en) * | 2018-08-20 | 2019-01-01 | 深圳壹账通智能科技有限公司 | Method, apparatus, terminal and the computer readable storage medium of recognition of face |
| CN109583972A (en) * | 2018-12-19 | 2019-04-05 | 中新智擎科技有限公司 | A kind of advertisement demonstration method, device, advertisement robot and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324912A (en) * | 2013-05-30 | 2013-09-25 | 苏州福丰科技有限公司 | Face recognition system and method for ATM |
| EP2833294A2 (en) * | 2013-07-30 | 2015-02-04 | Fujitsu Limited | Device to extract biometric feature vector, method to extract biometric feature vector and program to extract biometric feature vector |
| CN104573652A (en) * | 2015-01-04 | 2015-04-29 | 华为技术有限公司 | Method, device and terminal for determining identity identification of human face in human face image |
| CN106022317A (en) * | 2016-06-27 | 2016-10-12 | 北京小米移动软件有限公司 | Face identification method and apparatus |
| CN106156755A (en) * | 2016-07-29 | 2016-11-23 | 深圳云天励飞技术有限公司 | Similarity calculating method in a kind of recognition of face and system |
-
2016
- 2016-12-16 CN CN201611169854.8A patent/CN106650658A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324912A (en) * | 2013-05-30 | 2013-09-25 | 苏州福丰科技有限公司 | Face recognition system and method for ATM |
| EP2833294A2 (en) * | 2013-07-30 | 2015-02-04 | Fujitsu Limited | Device to extract biometric feature vector, method to extract biometric feature vector and program to extract biometric feature vector |
| CN104573652A (en) * | 2015-01-04 | 2015-04-29 | 华为技术有限公司 | Method, device and terminal for determining identity identification of human face in human face image |
| CN106022317A (en) * | 2016-06-27 | 2016-10-12 | 北京小米移动软件有限公司 | Face identification method and apparatus |
| CN106156755A (en) * | 2016-07-29 | 2016-11-23 | 深圳云天励飞技术有限公司 | Similarity calculating method in a kind of recognition of face and system |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108492575A (en) * | 2018-04-11 | 2018-09-04 | 济南浪潮高新科技投资发展有限公司 | A kind of intelligent vehicle type identifier method |
| CN109117801A (en) * | 2018-08-20 | 2019-01-01 | 深圳壹账通智能科技有限公司 | Method, apparatus, terminal and the computer readable storage medium of recognition of face |
| CN109583972A (en) * | 2018-12-19 | 2019-04-05 | 中新智擎科技有限公司 | A kind of advertisement demonstration method, device, advertisement robot and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109670528B (en) | Data expansion method facing pedestrian re-identification task and based on paired sample random occlusion strategy | |
| CN107871100A (en) | The training method and device of faceform, face authentication method and device | |
| CN106022380A (en) | Individual identity identification method based on deep learning | |
| CN106709511A (en) | Urban rail transit panoramic monitoring video fault detection method based on depth learning | |
| CN104348829B (en) | A kind of network security situation sensing system and method | |
| CN109886141A (en) | A kind of pedestrian based on uncertainty optimization discrimination method again | |
| CN105426908B (en) | A kind of substation's attributive classification method based on convolutional neural networks | |
| CN108921206A (en) | A kind of image classification method, device, electronic equipment and storage medium | |
| CN107451510A (en) | Biopsy method and In vivo detection system | |
| CN106203242A (en) | A kind of similar image recognition methods and equipment | |
| CN107832835A (en) | The light weight method and device of a kind of convolutional neural networks | |
| CN103888541B (en) | Method and system for discovering cells fused with topology potential and spectral clustering | |
| CN110266528B (en) | Traffic prediction method for Internet of vehicles communication based on machine learning | |
| CN109558902A (en) | A kind of fast target detection method | |
| CN104883278A (en) | Method for classifying network equipment by utilizing machine learning | |
| CN106485325B (en) | Two phase flow multivariate information fusion method based on complex network and deep learning and application | |
| CN104809393B (en) | A kind of support attack detecting algorithm based on popularity characteristic of division | |
| CN108021940A (en) | data classification method and system based on machine learning | |
| CN106650658A (en) | Method and device for identifying user identity | |
| CN110533116A (en) | Based on the adaptive set of Euclidean distance at unbalanced data classification method | |
| CN108696453A (en) | The power telecom network SDN Business Streams cognitive method and system of lightweight | |
| CN108156114A (en) | The key node of power information physical system network attack map determines method and device | |
| CN110069959A (en) | A kind of method for detecting human face, device and user equipment | |
| CN107871103A (en) | Face authentication method and device | |
| CN111581445A (en) | Graph embedding learning method based on graph elements |
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 | ||
| CB02 | Change of applicant information |
Address after: 518000 Shekou Torch Pioneering Building, Nanshan District, Shenzhen City, Guangdong Province, 2nd Floor Applicant after: New wisdom Technology Co., Ltd. Address before: 518000 Shekou Torch Pioneering Building, Nanshan District, Shenzhen City, Guangdong Province, 2nd Floor Applicant before: Sino wisdom Machinery Co., Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170510 |
|
| RJ01 | Rejection of invention patent application after publication |