CN106599829A - Face anti-counterfeiting algorithm based on active near-infrared light - Google Patents
Face anti-counterfeiting algorithm based on active near-infrared light Download PDFInfo
- Publication number
- CN106599829A CN106599829A CN201611130364.7A CN201611130364A CN106599829A CN 106599829 A CN106599829 A CN 106599829A CN 201611130364 A CN201611130364 A CN 201611130364A CN 106599829 A CN106599829 A CN 106599829A
- Authority
- CN
- China
- Prior art keywords
- layers
- layer
- infrared light
- face
- light source
- 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
- G06V40/40—Spoof detection, e.g. liveness detection
- G06V40/45—Detection of the body part being alive
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of face recognition, and in particular to a face anti-counterfeiting algorithm based on active near-infrared light. The method includes the following steps: S1. acquiring positive samples of a face under a near-infrared light source; S2. acquiring negative samples of an image under the near-infrared light source; S3. using the SGD optimization method to train a classifier based on a deep neural network; and S4. using the classifier to conduct face recognition to obtain a result. Compared with prior art, according to the application, the method has the following advantages: 1. capability of conducting bio-assay; 2. high classifying precision; 3. fast classifying speed; and 4. no susceptibility to visible light due to utilization of the near-infrared light source.
Description
Technical field
The present invention relates to field of face identification, more particularly to a kind of anti-pseudo Algorithm of face based on active near infrared light.
Background technology
At present, in image procossing and area of pattern recognition, obtained by the artificial neural network of manual manual features and shallow-layer
The feature for taking is being classified and be recognized.Under complicated environmental condition, these shallow-layer features are inadequate for identification.Deep layer
Neutral net be that deep learning is arisen at the historic moment, image and area of pattern recognition have been widely applied.
The basic procedure of depth model training (i.e. depth network training, deep learning systematic training) is briefly described below.
Every layer parameter of network is all expressed as (w, b) substantially, and wherein w is weighting parameter, and b is offset parameter, and per layer of input and output are closed
System is y=wx+b, wherein, x represents input, and y represents output.It is exactly a nest relation that each layer is coupled together, and is simple meter, false
Fixed total parameter is (W, B), and total input/output relation is Y=F (X, W, B).If model is trained, i.e., (W, B) is true
It is fixed, then it is exactly required result to output Y before having input X to directly obtain.If model is not also trained, i.e., (W, B) is without true
It is fixed, then (W, B) initial value (W0, B0) is first given, prediction output Y0=F (X, W0, B0) of training sample is obtained, it and training
The label of sample demarcates output Ytrue and there is very big deviation.One loss function can be set, such as loss=0.5*
(Ytrue-Y0) ^2, i.e. prediction output and label differ more remote, then loss function is bigger, at this moment carries out error-duration model to update mould
Shape parameter.Often train once, just update parameter (W, B) once, its purpose is just so that prediction output and demarcates the difference of output
Value is less and less, through the multiple training of many training samples, when loss values are less than certain value, is considered as model training good
(have found suitable (W, B) value), training process terminates.
But the existing recognition of face based on simulated training mainly realizes recognition of face according to picture, but can not be real
Existing In vivo detection.In other words, prior art is all to the recognition result of real face and the human face photo of same person
Success, thus causes potential safety hazard.In addition, existing grader precision not enough, needs further raising.
The content of the invention
It is an object of the invention to avoid the deficiency existing for above-mentioned prior art, propose a kind of based on active near-infrared
The anti-pseudo Algorithm of face of light, it can solve the problem that the problem that cannot realize In vivo detection.
The invention provides a kind of anti-pseudo Algorithm of the face based on active near infrared light, comprises the following steps：
S1：The positive class sample of face is gathered under near-infrared light source；
S2：The negative class sample of picture is gathered under near-infrared light source；
S3：A grader based on deep neural network is trained using SGD optimization methods；
S4：Recognition of face is carried out using grader, result is obtained.
Further, in step S3 CNN graders include be sequentially connected ground floor, the second layer, third layer, the 4th layer,
Layer 5, layer 6, the ground floor is Input layers, and the second layer includes Conv0 layers, Bn0 layers, the Relu0 being sequentially connected
Layer and Pool0 layers, the third layer include be sequentially connected Conv1 layers, Bn1 layers, Relu1 layers, Conv2 layers, Bn2 layers,
Relu2 layers, Pool1 layers, described 4th layer includes the Local0 layers, Bn3 layers, Relu3 layers, the Pool2 layers that are sequentially connected, described the
Five layers of Local1 layers, Bn4 layers, AvgPool layers for including being sequentially connected, the layer 6 is Softmax layers.
Further, SGD optimization methods are in step S2：For training sample set, we are divided into first n
Batch, each batch include m sample.Our each more new capital utilize the data of a batch, rather than whole training set,
I.e.：
xt+1=xt+Δxt
Δxt=-η gt
Wherein, η is learning rate, gtFor x t gradient.
Compared with prior art, the application has advantages below：1. In vivo detection can be realized；2. nicety of grading is high；3.
Classification speed is fast；4. near-infrared light source is adopted, it is not influenced by visible light.
Description of the drawings
With reference to the accompanying drawings and detailed description the present invention is further detailed explanation.
Fig. 1 is the schematic flow sheet of the embodiment of the present invention 1；
Fig. 2 is the structural representation of grader in the embodiment of the present invention 1.
Specific embodiment
Below in conjunction with accompanying drawing, technical scheme is further described, but the present invention is not limited to these realities
Apply example.
In this application, there is provided a kind of anti-pseudo Algorithm of the face based on active near infrared light.Because the face of live body is tied
Structure is three-dimensional, and the structure of photo is plane, and under near-infrared light source, the face of stereochemical structure shows bright place and the moon
At shadow, so the bright dark distribution of the living body faces for collecting can be different from the face on photo.Based on this, we will collect
The sample of living body faces as positive class sample, and using the picture sample for collecting as negative class sample.Further according to positive class sample and
Negative class sample, trains a grader based on deep neural network, so as to realize In vivo detection, people using SGD optimization methods
Face is recognized.It is described in detail in the following embodiments.
With reference to Fig. 1, the present embodiment is comprised the following steps：
S1：The positive class sample of face is gathered under near-infrared light source；
S2：The negative class sample of picture is gathered under near-infrared light source；
S3：A grader based on deep neural network is trained using SGD optimization methods
S4：Recognition of face is carried out using grader, result is obtained.
With reference to accompanying drawing 2, CNN graders include being sequentially connected in step S3 ground floor, the second layer, third layer, the 4th layer,
Layer 5, layer 6, the ground floor is Input layers, and the second layer includes Conv0 layers, Bn0 layers, the Relu0 being sequentially connected
Layer and Pool0 layers, the third layer include be sequentially connected Conv1 layers, Bn1 layers, Relu1 layers, Conv2 layers, Bn2 layers,
Relu2 layers, Pool1 layers, described 4th layer includes the Local0 layers, Bn3 layers, Relu3 layers, the Pool2 layers that are sequentially connected, described the
Five layers of Local1 layers, Bn4 layers, AvgPool layers for including being sequentially connected, the layer 6 is Softmax layers.
Input picture, picture size is generally 56x56x1；Into the second layer, carry out being entered after convolution into Conv0 layers
Bn0 layers will finally enter pool layer down-samplings after data normalization to the variance of zero-mean one into Relu0 layer activation primitives；Into
Third layer, Bn1 layers are entered into after Conv1 layer convolution will be activated after data normalization to the variance of zero-mean one into Relu1 layers
Function, Bn2 layers is entered into after Conv2 convolution and Relu2 layer activation primitives will be entered after data normalization to the variance of zero-mean one,
Finally enter pool1 layer down-samplings；Into the 4th layer, Bn3 layers are entered into after Local0 layers by data normalization to zero-mean
Relu3 layer activation primitives are entered after one variance, pool2 layer down-samplings are finally entered；It is laggard into Local1 layers into layer 5
Enter Bn4 layers by data normalization to the variance of zero-mean one, finally enter the mean value calculation after AvgPool layers are sampled；Most
Layer 6 Softmax layers are entered afterwards.
Further, SGD optimization methods are in step S2：For training sample set, we are divided into first n
Batch, each batch include m sample.Our each more new capital utilize the data of a batch, rather than whole training set,
I.e.：
xt+1=xt+Δxt
Δxt=-η gt
Wherein, η is learning rate, gtFor x t gradient.
Do so and be advantageous in that：When training data is too many, is updated using whole data set and do not shown on the often time.
The method of batch can reduce the pressure of machine, and can quickly restrain.It is (similar when training set has many redundancies
Sample occurs multiple), batch methods restrain faster.By taking an extreme case as an example, if training set the first half and later half gradient
It is identical.If that the first half is used as a batch, later half is used as another batch, then once traveling through training set
When, the method for batch to optimal solution is advanced two step, and the method for entirety is only advanced a step.General training data
For 10w positive and negative samples pictures, the degree of accuracy of identification is 99%, wherein, training parameter lr=0.01.
Specific embodiment described herein is only explanation for example spiritual to the present invention.Technology neck belonging to of the invention
The technical staff in domain can be made various modifications to described specific embodiment or supplement or replaced using similar mode
Generation, but without departing from the spiritual of the present invention or surmount scope defined in appended claims.
Claims (3)
1. the anti-pseudo Algorithm of face of active near infrared light is based on, it is characterised in that comprised the following steps：
S1：The positive class sample of face is gathered under near-infrared light source；
S2：The negative class sample of picture is gathered under near-infrared light source；
S3：A grader based on deep neural network is trained using SGD optimization methods；
S4：Recognition of face is carried out using grader, result is obtained.
2. the anti-pseudo Algorithm of the face based on active near infrared light according to claim 1, it is characterised in that in step S3 point
Class device includes ground floor, the second layer, third layer, the 4th layer, layer 5, the layer 6 being sequentially connected, and the ground floor is Input
Layer, the second layer includes the Conv0 layers, Bn0 layers, Relu0 layers and the Pool0 layers that are sequentially connected, the third layer include according to
The Conv1 layers of secondary connection, Bn1 layers, Relu1 layers, Conv2 layers, Bn2 layers, Relu2 layers, Pool1 layers, described 4th layer include according to
The Local0 layers of secondary connection, Bn3 layers, Relu3 layers, Pool2 layers, the layer 5 includes Local1 layers, the Bn4 being sequentially connected
Layer, AvgPool layers, the layer 6 is Softmax layers.
3. the anti-pseudo Algorithm of the face based on active near infrared light according to claim 1, it is characterised in that in step S2
SGD optimization methods are：For training sample set, we are divided into first n batch, and each batch includes m sample, I
Every time more new capital using a batch data, rather than whole training set, i.e.,：
xt+1=xt+Δxt
Δxt=-η gt
Wherein, η is learning rate, gtFor x t gradient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611130364.7A CN106599829A (en) | 2016-12-09 | 2016-12-09 | Face anti-counterfeiting algorithm based on active near-infrared light |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611130364.7A CN106599829A (en) | 2016-12-09 | 2016-12-09 | Face anti-counterfeiting algorithm based on active near-infrared light |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106599829A true CN106599829A (en) | 2017-04-26 |
Family
ID=58597901
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611130364.7A Pending CN106599829A (en) | 2016-12-09 | 2016-12-09 | Face anti-counterfeiting algorithm based on active near-infrared light |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106599829A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107590473A (en) * | 2017-09-19 | 2018-01-16 | 杭州登虹科技有限公司 | A kind of human face in-vivo detection method, medium and relevant apparatus |
| CN107609463A (en) * | 2017-07-20 | 2018-01-19 | 百度在线网络技术（北京）有限公司 | Biopsy method, device, equipment and storage medium |
| CN108363944A (en) * | 2017-12-28 | 2018-08-03 | 杭州宇泛智能科技有限公司 | Recognition of face terminal is double to take the photograph method for anti-counterfeit, apparatus and system |
| CN108776786A (en) * | 2018-06-04 | 2018-11-09 | 北京京东金融科技控股有限公司 | Method and apparatus for generating user's truth identification model |
| CN108875467A (en) * | 2017-06-05 | 2018-11-23 | 北京旷视科技有限公司 | The method, apparatus and computer storage medium of In vivo detection |
| CN108985134A (en) * | 2017-06-01 | 2018-12-11 | 重庆中科云丛科技有限公司 | Face In vivo detection and brush face method of commerce and system based on binocular camera |
| CN109255322A (en) * | 2018-09-03 | 2019-01-22 | 北京诚志重科海图科技有限公司 | A kind of human face in-vivo detection method and device |
| CN109376595A (en) * | 2018-09-14 | 2019-02-22 | 杭州宇泛智能科技有限公司 | Monocular RGB camera in-vivo detection method and system based on human eye attention |
| CN110738072A (en) * | 2018-07-18 | 2020-01-31 | 浙江宇视科技有限公司 | Living body judgment method and device |
| CN110969189A (en) * | 2019-11-06 | 2020-04-07 | 杭州宇泛智能科技有限公司 | Face detection method and device and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024145A (en) * | 2010-12-01 | 2011-04-20 | 五邑大学 | Layered recognition method and system for disguised face |
| CN105069448A (en) * | 2015-09-29 | 2015-11-18 | 厦门中控生物识别信息技术有限公司 | True and false face identification method and device |
| CN105608450A (en) * | 2016-03-01 | 2016-05-25 | 天津中科智能识别产业技术研究院有限公司 | Heterogeneous face identification method based on deep convolutional neural network |
-
2016
- 2016-12-09 CN CN201611130364.7A patent/CN106599829A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024145A (en) * | 2010-12-01 | 2011-04-20 | 五邑大学 | Layered recognition method and system for disguised face |
| CN105069448A (en) * | 2015-09-29 | 2015-11-18 | 厦门中控生物识别信息技术有限公司 | True and false face identification method and device |
| CN105608450A (en) * | 2016-03-01 | 2016-05-25 | 天津中科智能识别产业技术研究院有限公司 | Heterogeneous face identification method based on deep convolutional neural network |
Non-Patent Citations (3)
| Title |
|---|
| JIANWEI YANG等: "Learn Convolutional Neural Network for Face Anti-Spoofing", 《COMPUTER SCIENCE》 * |
| MIN PENG等: "NIRFaceNet: A Convolutional Neural Network for Near-Infrared Face Identification", 《INFORMATION》 * |
| 刘旷: "基于卷积网络集成的面部表情识别方法", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108985134A (en) * | 2017-06-01 | 2018-12-11 | 重庆中科云丛科技有限公司 | Face In vivo detection and brush face method of commerce and system based on binocular camera |
| CN108875467A (en) * | 2017-06-05 | 2018-11-23 | 北京旷视科技有限公司 | The method, apparatus and computer storage medium of In vivo detection |
| CN108875467B (en) * | 2017-06-05 | 2020-12-25 | 北京旷视科技有限公司 | Living body detection method, living body detection device and computer storage medium |
| CN107609463A (en) * | 2017-07-20 | 2018-01-19 | 百度在线网络技术（北京）有限公司 | Biopsy method, device, equipment and storage medium |
| CN107609463B (en) * | 2017-07-20 | 2021-11-23 | 百度在线网络技术（北京）有限公司 | Living body detection method, living body detection device, living body detection equipment and storage medium |
| CN107590473A (en) * | 2017-09-19 | 2018-01-16 | 杭州登虹科技有限公司 | A kind of human face in-vivo detection method, medium and relevant apparatus |
| CN108363944A (en) * | 2017-12-28 | 2018-08-03 | 杭州宇泛智能科技有限公司 | Recognition of face terminal is double to take the photograph method for anti-counterfeit, apparatus and system |
| CN108776786A (en) * | 2018-06-04 | 2018-11-09 | 北京京东金融科技控股有限公司 | Method and apparatus for generating user's truth identification model |
| CN110738072A (en) * | 2018-07-18 | 2020-01-31 | 浙江宇视科技有限公司 | Living body judgment method and device |
| CN109255322B (en) * | 2018-09-03 | 2019-11-19 | 北京诚志重科海图科技有限公司 | A kind of human face in-vivo detection method and device |
| CN109255322A (en) * | 2018-09-03 | 2019-01-22 | 北京诚志重科海图科技有限公司 | A kind of human face in-vivo detection method and device |
| CN109376595A (en) * | 2018-09-14 | 2019-02-22 | 杭州宇泛智能科技有限公司 | Monocular RGB camera in-vivo detection method and system based on human eye attention |
| CN110969189A (en) * | 2019-11-06 | 2020-04-07 | 杭州宇泛智能科技有限公司 | Face detection method and device and electronic equipment |
| CN110969189B (en) * | 2019-11-06 | 2023-07-25 | 杭州宇泛智能科技有限公司 | Face detection method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106599829A (en) | Face anti-counterfeiting algorithm based on active near-infrared light | |
| CN106682704B (en) | A kind of disease geo-radar image recognition methods of integrating context information | |
| CN106203318B (en) | Camera network pedestrian recognition method based on the fusion of multi-level depth characteristic | |
| CN107451607B (en) | A kind of personal identification method of the typical character based on deep learning | |
| CN104616664B (en) | A kind of audio identification methods detected based on sonograph conspicuousness | |
| CN108537191B (en) | Three-dimensional face recognition method based on structured light camera | |
| CN110210563A (en) | The study of pattern pulse data space time information and recognition methods based on Spike cube SNN | |
| CN107016406A (en) | The pest and disease damage image generating method of network is resisted based on production | |
| CN109034044A (en) | A kind of pedestrian's recognition methods again based on fusion convolutional neural networks | |
| CN107154048A (en) | The remote sensing image segmentation method and device of a kind of Pulse-coupled Neural Network Model | |
| CN109598891B (en) | A kind of method and system for realizing Smoke Detection using deep learning disaggregated model | |
| CN107229929A (en) | A kind of license plate locating method based on R CNN | |
| CN107248150A (en) | A kind of Multiscale image fusion methods extracted based on Steerable filter marking area | |
| CN107967685A (en) | A kind of bridge pier and tower crack harmless quantitative detection method based on unmanned aerial vehicle remote sensing | |
| CN106022380A (en) | Individual identity identification method based on deep learning | |
| CN108898085A (en) | A kind of road disease intelligent detecting method based on mobile video | |
| CN107085696A (en) | A kind of vehicle location and type identifier method based on bayonet socket image | |
| CN107992836A (en) | A kind of recognition methods of miner's unsafe acts and system | |
| CN108053398A (en) | A kind of melanoma automatic testing method of semi-supervised feature learning | |
| CN105825235A (en) | Image identification method based on deep learning of multiple characteristic graphs | |
| CN103914699A (en) | Automatic lip gloss image enhancement method based on color space | |
| CN104077579A (en) | Facial expression image recognition method based on expert system | |
| CN104732494A (en) | Tissue culturing monitoring method and system based on image mist elimination | |
| CN110675462A (en) | Gray level image colorizing method based on convolutional neural network | |
| CN108460399A (en) | A kind of child building block builds householder method and system |
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 |
Application publication date: 20170426 |
|
| RJ01 | Rejection of invention patent application after publication |