CN109409342A - A kind of living iris detection method based on light weight convolutional neural networks - Google Patents
A kind of living iris detection method based on light weight convolutional neural networks Download PDFInfo
- Publication number
- CN109409342A CN109409342A CN201811511287.9A CN201811511287A CN109409342A CN 109409342 A CN109409342 A CN 109409342A CN 201811511287 A CN201811511287 A CN 201811511287A CN 109409342 A CN109409342 A CN 109409342A
- Authority
- CN
- China
- Prior art keywords
- iris
- iris image
- neural networks
- convolutional neural
- light weight
- 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
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/193—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/197—Matching; Classification
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Ophthalmology & Optometry (AREA)
- Probability & Statistics with Applications (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Image Analysis (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
A kind of living iris detection method based on light weight convolutional neural networks disclosed by the invention, include: S1, acquisition living body iris image, prosthese iris image sample, obtains distinguishing the network weight and classifier of living body iris and prosthese iris image through convolutional neural networks learning training；S2, iris image to be detected is obtained；S3, feature extraction is carried out to iris image to be detected using trained light weight convolutional neural networks model, obtains the corresponding feature vector of iris image to be detected；S4, the corresponding feature vector of iris image to be measured that S3 is obtained is input in the classifier that S1 is obtained, calculate the probability value that described eigenvector belongs to living body iris image, when the probability value be more than default threshold value, that is, can determine that iris image to be detected be living body.It is an advantage of the current invention that iris equipment is avoided " to capture " problem by prosthese iris, so that iris identification equipment has higher reliability and safety, the confidence that user uses iris identification equipment is increased.
Description
Technical field
The present invention relates to image identification technical field, specially a kind of iris living body inspection based on light weight convolutional neural networks
Survey method.
Background technique
With the development of science and technology, biometrics identification technology is widely used in many industries, due to iris
Identification technology has the characteristics that high stability, high antifalsification and uniqueness, it is made to be widely used, and especially those maintain secrecy
In the more demanding scene of property.But with popularizing for biometrics identification technology, iris is forged technology and is also occurred therewith.Therefore,
Whether iris identification equipment is equipped with the horizontal height that In vivo detection technology embodies the equipment.
Iris In vivo detection technology at this stage, primarily focuses on the help of hardware, different strong by hardware device release
Weak light makes pupil that different contraction situations be presented, and carries out In vivo detection with this.But this method equipment manufacturing cost is higher, and
The light intensity period is discharged according to equipment, can forge pupil scaling interval reduces to can not judge whether iris is living body
The safety of safety-security area.
Summary of the invention
In order to solve the help that iris In vivo detection in the prior art lays particular emphasis on hardware, can forge pupil scale distance from
And the problem of can not truly judging whether it is living body, the present invention provides a kind of, and the iris based on light weight convolutional neural networks is living
The purpose of body detecting method, realization is to use whether light weight convolutional neural networks predict for living body iris data to be checked
And judgement, detection speed is fast, accuracy rate is high.
To achieve the goals above, technical solution provided by the invention is that one kind provided by the invention is based on light weight convolution
The living iris detection method of neural network, comprising:
S1, acquisition living body iris image, prosthese iris image sample are obtained distinguishing and be lived through convolutional neural networks learning training
The network weight and classifier of body iris and prosthese iris image；
S2, iris image to be detected is obtained；
S3, feature extraction is carried out to iris image to be detected using trained light weight convolutional neural networks model, obtained
The corresponding feature vector of iris image to be detected；
S4, the corresponding feature vector of iris image to be measured that S3 is obtained is input in the classifier that S1 is obtained, calculates institute
State the probability value that feature vector belongs to living body iris image, when the probability value be more than default threshold value, that is, can determine that be checked
The iris image of survey is living body.
Further, network weight described in S1 and classifier training process are as follows:
S11, trained iris image is treated first handled, including label, training set, verifying collection and test
The division of collection；
S12, secondly, being configured to the hyper parameter used in training process and the selection of optimizer algorithm；
S13, in the training process, data obtain the prediction result of each classification by the forward inference of network, utilize damage
The error between function calculating predicted value and true value is lost, then by error back propagation into network；
S14, in back-propagation process, local derviation is asked to each node, using learning rate as step-length adjust weight size, then
Secondary propagated forward, calculates repeatedly, when training to loss function value no longer changes or in a certain value slightly concussion up and down, stops
Only train；If classifier precision on verifying collection and test set is optimal at this time, model training is completed.
It further, further include using trained light weight convolutional neural networks model in the S3 to iris figure to be detected
As being pre-processed for the iris image, with the iris portion in the determination iris image before carrying out feature extraction.
It is described to be pre-processed for the iris data, comprising:
The iris portion of the iris image is positioned, and is split, to obtain circular iris region；
Polar coordinate transform is first done to the iris region of the annular shape, transform into it is rectangular-shaped, then from being divided into two among it,
And right part is copied to the lower section of left-side images to, then zoom in and out to it and normalized.According to following formula to contracting
Image after putting is normalized,
Wherein, x is the iris image pixel point value after the scaling of input, and the pixel value after x ' calculating, μ and σ are respectively to contract
Put expectation and the variance yields of rear iris image.
Further, iris image textural characteristics rich in, the differentiation of true and false body iris mainly according to its texture,
The distribution of gray scale is different.Feature is carried out to the iris pattern using trained light weight convolutional neural networks model in the S3
It extracts, obtains one group of feature vector to be sorted.
The neural network trunk model includes two convolutional layers, two Stage, a mean value pond layer and a classification
Layer.
Iris image is inputted in the S4, forward inference is carried out to data using trained weight, obtains one group
Feature, then classified by classifier, the probability value which belongs to a certain classification is obtained by following formula, if the probability is greater than
Given threshold then belongs to living body, otherwise is prosthese；
To sum up, the present invention by adopting the above technical scheme, has the advantages that, the present invention uses the convolution mind with light weight
In vivo detection is carried out to iris data through network, speed is fast for detection, accuracy rate is high, therefore the present invention can avoid iris equipment by
Prosthese iris " captures " problem, so that iris identification equipment has higher reliability and safety, increases user to iris
The confidence that identification equipment uses.
Detailed description of the invention
Fig. 1 is in S3 step of the present invention to schematic diagram before iris image normalized to be measured；
Fig. 2 is in S3 step of the present invention to schematic diagram after iris image normalized to be measured；
Fig. 3 is light weight neural network model backbone network structural schematic diagram of the present invention；
In figure, 1, iris image；2, outer circle；3, inner circle；4, circular iris region；5, rectangular-shaped iris region；6, it adjusts
Iris image after whole；7, the iris image after scaling and normalize.
Specific embodiment
The present invention is further illustrated below in conjunction with specific embodiment.
A kind of embodiment one: living iris detection method based on light weight convolutional neural networks provided by the invention, comprising:
S1, acquisition living body iris image, prosthese iris image sample are obtained distinguishing and be lived through convolutional neural networks learning training
The network weight and classifier of body iris and prosthese iris image；
S2, iris image to be detected is obtained；
S3, feature extraction is carried out to iris image to be detected using trained light weight convolutional neural networks model, obtained
The corresponding feature vector of iris image to be detected；
S4, the corresponding feature vector of iris image to be measured that S3 is obtained is input in the classifier that S1 is obtained, calculates institute
State the probability value that feature vector belongs to living body iris image, when the probability value be more than default threshold value, that is, can determine that be checked
The iris image of survey is living body.
Further, network weight described in S1 and classifier training process are as follows:
S11, trained iris image is treated first handled, including label, training set, verifying collection and test
The division of collection；
S12, secondly, being configured to the hyper parameter used in training process and the selection of optimizer algorithm；
S13, in the training process, data obtain the prediction result of each classification by the forward inference of network, utilize damage
The error between function calculating predicted value and true value is lost, then by error back propagation into network；
S14, in back-propagation process, local derviation is asked to each node, using learning rate as step-length adjust weight size, then
Secondary propagated forward, calculates repeatedly, when training to loss function value no longer changes or in a certain value slightly concussion up and down, stops
Only train；If classifier precision on verifying collection and test set is optimal at this time, model training is completed.
It further, further include using trained light weight convolutional neural networks model in the S3 to iris to be detected
Before image carries out feature extraction, pre-processed for the iris image, with the iridial part of retina in the determination iris image
Point.It is described to be pre-processed for the iris data, comprising:
The iris portion of the iris image is positioned, and is split, to obtain circular iris region；
Polar coordinate transform is first done to the iris region of the annular shape, transform into it is rectangular-shaped, then from being divided into two among it,
And right part is copied to the lower section of left-side images to, then zoom in and out to it and normalized.According to following formula to contracting
Image after putting is normalized,
Wherein, x is the iris image pixel point value after the scaling of input, and the pixel value after x ' calculating, μ and σ are respectively to contract
Put expectation and the variance yields of rear iris image.
Being referred to the original image shown in Fig. 1, got is iris image 1 (left figure), and there are outer circles 2 and interior in the figure
Circle 3 positions and determines part of the iris portion between outer circle 2 and inner circle 3 in iris image 1, therefore carries out for iris image 1
Segmentation obtains circular iris region 4, i.e. right figure dash area in Fig. 1.The circular iris region of control carries out polar coordinates
Conversion process obtains rectangular-shaped iris region 5 as shown in Figure 2.Iris region will be held to be divided into two from centre, by right side
An image 6 is formed below copied part to left-side images, then zooms to pre-set dimension 7, then be sent into after normalized
Feature extraction is carried out in neural network.
In the present invention, by the pretreatment to iris image, only iris region part is obtained, is then fed into network
Training and test, effectively avoid interference of the non-iris region to classification task, while can also accelerate net in the training process
The convergent speed of network.
Further, the iris pattern is carried out using trained light weight convolutional neural networks model in the S3
When feature extraction, the light weight neural network model is referring to shown in Fig. 3:
The neural network trunk model includes two convolutional layers, two Stage, a mean value pond layer and a classification
Two convolutional layer of layer are all common convolutional layer, and second convolutional layer uses 1 × 1 convolution kernel, and main purpose is to use
Lesser calculation amount improves the ability to express of feature；Two Stage layers are made of Fig. 3 (a) and (b), and wherein Fig. 3 (a) is
The main composition of Stage, Fig. 3 (b) are the purposes for playing down-sampling.
In Fig. 3 (a), Channel Splite is that input data channel is divided into two parts, and a part is directly transmitted
To Concat layers, a part passes to Concat layers after 3 convolutional layers are handled again and is linked, and the step-length of each convolutional layer is
1, Channel layers of Shuffle are then passed to, channel is carried out to characteristic pattern and is shuffled, the ability to express of feature is improved.
Unlike Fig. 3 (a), more down-sampling processes, lower using mainly by one 3 × 3 on the left of Fig. 3 (b)
DepthWise layer and one 1 × 1 PointWise layer composition, at this timeDepthWise layersStep-length is set as 2, the 3 of right side
× 3 DepthWise layer step-length is also 2, and other parts are identical.Down-sampling is only done in the first time of each Stage, other 5 times
It is executed using Fig. 3 (a) structure.
3, right side convolutional layer main composition in Fig. 3 (a) has, and first convolution kernel is 1 × 1, in traditional BottleNeck
In layer, which is to play the role of dimensionality reduction, but in network in the present invention, the output layer of this layer and discrepancy layer characteristic pattern channel
Number is identical, and main purpose is to improve network reasoning speed to reduce internal storage access cost；Second is respectively with third
DepthWise and PointWise layers, it is therefore an objective to for the parameter amount and calculation amount for reducing network.
The size of the input data of different neural network models requires, for example, neural network in the present embodiment
Input data requirement is 48 × 48 sizes.When practical application, rectangular-shaped iris region as shown in Figure 2 can be handled into pressure
It is condensed to the image that pixel is 48 × 48.That is, the present embodiment carries out a series of place by the iris region to ring-shaped
Reason, finally obtains pre-set dimension image, to meet the input requirements of neural network model.
Iris image is inputted in the S4, forward inference is carried out to data using trained weight, obtains one group
Feature, then classified by classifier, the probability value which belongs to a certain classification is obtained by following formula, if the probability is greater than
Given threshold then belongs to living body, otherwise is prosthese；
Following tables 1 is whole network framework of the invention, is had by the model that the network architecture trains to iris image
Preferable prediction classifying quality can further determine whether input iris image is true body according to classification results.
Table 1
| Layer | Output Size | Ksize | Stride | Repeat | Output Channels |
| Image | 48×48 | 1 | |||
| Conv1 | 24×24 | 3×3 | 2 | 1 | 32 |
| Stage1 | 12×12 | 2/1 | 6 | 176 | |
| Stage2 | 6×6 | 2/1 | 6 | 352 | |
| Conv4 | 6×6 | 1×1 | 1 | 1 | 512 |
| GlobalPool | 1×1 | 6×6 | |||
| FC | 2 |
The present invention, it is special to be carried out using trained neural network model to iris image by obtaining iris image
Sign is extracted, and the corresponding feature vector of iris image is obtained, and provides preparation to calculate the similarity of feature vector and standard picture, real
The purpose that the corresponding image classification of iris image is determined now depending on similarity is solved and is carried out in the prior art according to iris image
The image classification dependence excessive to hardware device, realize improved by convolutional neural networks to the accuracy of iris classification and
Shandong nation property.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (6)
1. a kind of living iris detection method based on light weight convolutional neural networks characterized by comprising
S1, acquisition living body iris image, prosthese iris image sample, obtain distinguishing living body rainbow through convolutional neural networks learning training
The network weight and classifier of film and prosthese iris image；
S2, iris image to be detected is obtained；
S3, feature extraction is carried out to iris image to be detected using trained light weight convolutional neural networks model, obtained to be checked
Survey the corresponding feature vector of iris image；
S4, the corresponding feature vector of iris image to be measured that S3 is obtained is input in the classifier that S1 is obtained, calculates the spy
Sign vector belongs to the probability value of living body iris image, when gained probability value is more than default threshold value, that is, can determine that be detected
Iris image is living body.
2. the living iris detection method according to claim 1 based on light weight convolutional neural networks, which is characterized in that S1
The network weight and classifier training process are as follows:
S11, trained iris image is treated first handled, including label, training set, verifying collection and test set
It divides；
S12, secondly, being configured to the hyper parameter used in training process and the selection of optimizer algorithm；
S13, in the training process, data obtain the prediction result of each classification by the forward inference of network, utilize loss letter
Number calculates the error between predicted value and true value, then by error back propagation into network；
S14, in back-propagation process, local derviation is asked to each node, using learning rate as step-length adjust weight size, again before
It to propagation, calculates repeatedly, when training to loss function value no longer changes or in a certain value slightly concussion up and down, stops instruction
Practice；If classifier precision on verifying collection and test set is optimal at this time, model training is completed.
3. the living iris detection method according to claim 1 based on light weight convolutional neural networks, which is characterized in that institute
State in S3 further include using trained light weight convolutional neural networks model to iris image to be detected carry out feature extraction before,
It is pre-processed for the iris image, with the iris portion in the determination iris image.
4. the living iris detection method according to claim 3 based on light weight convolutional neural networks, which is characterized in that institute
It states and is pre-processed for the iris data, comprising:
The iris portion of the iris image is positioned, and is split, to obtain circular iris region；
Polar coordinate transform is first done to the iris region of the annular shape, transform into it is rectangular-shaped, then from being divided into two among it, and will
Right part copies the lower section of left-side images to, then zooms in and out to it and normalized；According to following formula to scaling after
Image be normalized,
Wherein, x is the iris image pixel point value after the scaling of input, and the pixel value after x ' calculating, μ and σ are respectively after scaling
The expectation of iris image and variance yields.
5. the living iris detection method according to claim 1 or 4 based on light weight convolutional neural networks, feature exist
In,
Iris image textural characteristics rich in, true and false body iris differentiation mainly according to its texture, gray scale distribution not
Together；Feature extraction is carried out to the iris pattern using trained light weight convolutional neural networks model in the S3, will be obtained
One group of feature vector to be sorted；
The light weight neural network trunk model includes two convolutional layers, two Stage layers, a mean value pond layer and one point
Class layer.
6. the living iris detection method according to claim 1 based on light weight convolutional neural networks, which is characterized in that
Iris image is inputted in the S4, forward inference is carried out to data using trained weight, obtains one group of feature,
Classified again by classifier, the probability value which belongs to a certain classification is obtained by following formula, if the probability is greater than setting
Threshold value then belongs to living body, otherwise is prosthese；
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811511287.9A CN109409342A (en) | 2018-12-11 | 2018-12-11 | A kind of living iris detection method based on light weight convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811511287.9A CN109409342A (en) | 2018-12-11 | 2018-12-11 | A kind of living iris detection method based on light weight convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109409342A true CN109409342A (en) | 2019-03-01 |
Family
ID=65458489
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811511287.9A Pending CN109409342A (en) | 2018-12-11 | 2018-12-11 | A kind of living iris detection method based on light weight convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109409342A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110046588A (en) * | 2019-04-22 | 2019-07-23 | 吉林大学 | It is a kind of with steal attack coping mechanism heterogeneous iris recognition method |
| CN110321844A (en) * | 2019-07-04 | 2019-10-11 | 北京万里红科技股份有限公司 | A kind of quick iris detection method based on convolutional neural networks |
| CN110909601A (en) * | 2019-10-18 | 2020-03-24 | 武汉虹识技术有限公司 | Beautiful pupil identification method and system based on deep learning |
| CN111027464A (en) * | 2019-12-09 | 2020-04-17 | 大连理工大学 | Iris identification method for convolutional neural network and sequence feature coding joint optimization |
| CN111242063A (en) * | 2020-01-17 | 2020-06-05 | 江苏大学 | Small sample classification model construction method based on transfer learning and iris classification application |
| CN111402143A (en) * | 2020-06-03 | 2020-07-10 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and computer readable storage medium |
| CN111401145A (en) * | 2020-02-26 | 2020-07-10 | 三峡大学 | Visible light iris recognition method based on deep learning and DS evidence theory |
| CN111507198A (en) * | 2020-03-25 | 2020-08-07 | 北京万里红科技股份有限公司 | Training method for printed iris detection model, and printed iris detection method and device |
| CN111611851A (en) * | 2020-04-10 | 2020-09-01 | 北京中科虹霸科技有限公司 | Model generation method, iris detection method and device |
| CN111798404A (en) * | 2019-04-04 | 2020-10-20 | 上海点与面智能科技有限公司 | Iris image quality evaluation method and system based on deep neural network |
| CN111914646A (en) * | 2020-07-01 | 2020-11-10 | 天津中科智能识别产业技术研究院有限公司 | Double-current fusion network iris in-vivo detection method based on light field image sequence |
| CN112927174A (en) * | 2019-12-06 | 2021-06-08 | 阿里巴巴集团控股有限公司 | Method and device for image processing and image training to channel shuffling |
| CN113837993A (en) * | 2021-07-29 | 2021-12-24 | 天津中科智能识别产业技术研究院有限公司 | Lightweight iris image segmentation method and device, electronic equipment and storage medium |
| CN113837117A (en) * | 2021-09-28 | 2021-12-24 | 上海电力大学 | Novel normalization and deep neural network-based iris coding method |
| CN115100730A (en) * | 2022-07-21 | 2022-09-23 | 北京万里红科技有限公司 | Iris living body detection model training method, iris living body detection method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101154265A (en) * | 2006-09-29 | 2008-04-02 | 中国科学院自动化研究所 | Method for recognizing iris with matched characteristic and graph based on partial bianry mode |
| CN101833646A (en) * | 2009-03-11 | 2010-09-15 | 中国科学院自动化研究所 | Living iris detection method |
| CN106326874A (en) * | 2016-08-30 | 2017-01-11 | 天津中科智能识别产业技术研究院有限公司 | Method and device for recognizing iris in human eye images |
| CN107220598A (en) * | 2017-05-12 | 2017-09-29 | 中国科学院自动化研究所 | Iris Texture Classification based on deep learning feature and Fisher Vector encoding models |
-
2018
- 2018-12-11 CN CN201811511287.9A patent/CN109409342A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101154265A (en) * | 2006-09-29 | 2008-04-02 | 中国科学院自动化研究所 | Method for recognizing iris with matched characteristic and graph based on partial bianry mode |
| CN101833646A (en) * | 2009-03-11 | 2010-09-15 | 中国科学院自动化研究所 | Living iris detection method |
| CN106326874A (en) * | 2016-08-30 | 2017-01-11 | 天津中科智能识别产业技术研究院有限公司 | Method and device for recognizing iris in human eye images |
| CN107220598A (en) * | 2017-05-12 | 2017-09-29 | 中国科学院自动化研究所 | Iris Texture Classification based on deep learning feature and Fisher Vector encoding models |
Non-Patent Citations (3)
| Title |
|---|
| NINGNING MA ET AL: "ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design", 《ARXIV:1807.11164V1》 * |
| 张传雷等: "《基于图像分析的植物及其病虫害识别方法研究》", 31 October 2018 * |
| 李志明: "基于卷积神经网络的虹膜活体检测算法研究", 《计算机工程》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111798404A (en) * | 2019-04-04 | 2020-10-20 | 上海点与面智能科技有限公司 | Iris image quality evaluation method and system based on deep neural network |
| CN110046588A (en) * | 2019-04-22 | 2019-07-23 | 吉林大学 | It is a kind of with steal attack coping mechanism heterogeneous iris recognition method |
| CN110321844B (en) * | 2019-07-04 | 2021-09-03 | 北京万里红科技股份有限公司 | Fast iris detection method based on convolutional neural network |
| CN110321844A (en) * | 2019-07-04 | 2019-10-11 | 北京万里红科技股份有限公司 | A kind of quick iris detection method based on convolutional neural networks |
| CN110909601A (en) * | 2019-10-18 | 2020-03-24 | 武汉虹识技术有限公司 | Beautiful pupil identification method and system based on deep learning |
| CN110909601B (en) * | 2019-10-18 | 2022-12-09 | 武汉虹识技术有限公司 | Beautiful pupil identification method and system based on deep learning |
| CN112927174B (en) * | 2019-12-06 | 2024-05-03 | 阿里巴巴集团控股有限公司 | Image processing, image training channel shuffling method and device |
| CN112927174A (en) * | 2019-12-06 | 2021-06-08 | 阿里巴巴集团控股有限公司 | Method and device for image processing and image training to channel shuffling |
| CN111027464A (en) * | 2019-12-09 | 2020-04-17 | 大连理工大学 | Iris identification method for convolutional neural network and sequence feature coding joint optimization |
| CN111027464B (en) * | 2019-12-09 | 2023-07-18 | 大连理工大学 | Iris recognition method for jointly optimizing convolutional neural network and sequence feature coding |
| CN111242063A (en) * | 2020-01-17 | 2020-06-05 | 江苏大学 | Small sample classification model construction method based on transfer learning and iris classification application |
| CN111242063B (en) * | 2020-01-17 | 2023-08-25 | 江苏大学 | Small sample classification model construction method based on transfer learning and iris classification application |
| CN111401145A (en) * | 2020-02-26 | 2020-07-10 | 三峡大学 | Visible light iris recognition method based on deep learning and DS evidence theory |
| CN111507198A (en) * | 2020-03-25 | 2020-08-07 | 北京万里红科技股份有限公司 | Training method for printed iris detection model, and printed iris detection method and device |
| CN111507198B (en) * | 2020-03-25 | 2023-11-28 | 北京万里红科技有限公司 | Training method for printing iris detection model, and printing iris detection method and device |
| CN111611851A (en) * | 2020-04-10 | 2020-09-01 | 北京中科虹霸科技有限公司 | Model generation method, iris detection method and device |
| CN111611851B (en) * | 2020-04-10 | 2024-04-02 | 北京中科虹霸科技有限公司 | Model generation method, iris detection method and device |
| CN111402143A (en) * | 2020-06-03 | 2020-07-10 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and computer readable storage medium |
| CN111914646A (en) * | 2020-07-01 | 2020-11-10 | 天津中科智能识别产业技术研究院有限公司 | Double-current fusion network iris in-vivo detection method based on light field image sequence |
| CN113837993A (en) * | 2021-07-29 | 2021-12-24 | 天津中科智能识别产业技术研究院有限公司 | Lightweight iris image segmentation method and device, electronic equipment and storage medium |
| CN113837993B (en) * | 2021-07-29 | 2024-01-30 | 天津中科智能识别产业技术研究院有限公司 | Lightweight iris image segmentation method and device, electronic equipment and storage medium |
| CN113837117A (en) * | 2021-09-28 | 2021-12-24 | 上海电力大学 | Novel normalization and deep neural network-based iris coding method |
| CN113837117B (en) * | 2021-09-28 | 2024-05-07 | 上海电力大学 | Iris coding method based on novel normalization and depth neural network |
| CN115100730A (en) * | 2022-07-21 | 2022-09-23 | 北京万里红科技有限公司 | Iris living body detection model training method, iris living body detection method and device |
| CN115100730B (en) * | 2022-07-21 | 2023-08-08 | 北京万里红科技有限公司 | Iris living body detection model training method, iris living body detection method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109409342A (en) | A kind of living iris detection method based on light weight convolutional neural networks | |
| CN110348319B (en) | Face anti-counterfeiting method based on face depth information and edge image fusion | |
| Adriman et al. | Performance evaluation of binary classification of diabetic retinopathy through deep learning techniques using texture feature | |
| CN107463920A (en) | A kind of face identification method for eliminating partial occlusion thing and influenceing | |
| Yao et al. | Robust CNN-based gait verification and identification using skeleton gait energy image | |
| CN111079518B (en) | Ground-falling abnormal behavior identification method based on law enforcement and case handling area scene | |
| CN105160317A (en) | Pedestrian gender identification method based on regional blocks | |
| CN111914761A (en) | Thermal infrared face recognition method and system | |
| CN110689526B (en) | Retinal blood vessel segmentation method and system based on retinal fundus image | |
| CN112580445B (en) | Human body gait image visual angle conversion method based on generation of confrontation network | |
| CN115131503A (en) | Health monitoring method and system for iris three-dimensional recognition | |
| Sujana et al. | An effective CNN based feature extraction approach for iris recognition system | |
| Yuan et al. | Ear detection based on CenterNet | |
| CN111144497B (en) | Image significance prediction method under multitasking depth network based on aesthetic analysis | |
| Liu et al. | DCMS-YOLOv5: A Dual-Channel and Multi-Scale Vertical Expansion Helmet Detection Model Based on YOLOv5. | |
| CN109815887B (en) | Multi-agent cooperation-based face image classification method under complex illumination | |
| Nouisser et al. | Enhanced MobileNet and transfer learning for facial emotion recognition | |
| CN103390150A (en) | Human body part detection method and device | |
| Méndez-Llanes et al. | On the use of local fixations and quality measures for deep face recognition | |
| CN104732521B (en) | A kind of similar purpose dividing method based on weight group similar active skeleton pattern | |
| US20230230277A1 (en) | Object position estimation device, object position estimation method, and recording medium | |
| Cai et al. | Multi-view joint learning network for pedestrian gender classification | |
| Liu et al. | Attention-Based feature pyramid network for object detection | |
| Zou et al. | Deep learning-based pavement cracks detection via wireless visible light camera-based network | |
| Arivazhagan et al. | Face recognition based on local directional number pattern and ANFIS classifier |
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 | ||
| CB02 | Change of applicant information |
Address after: 100081 Room 204, building 3, Fuhai center, Daliushu, Haidian District, Beijing Applicant after: Beijing wanlihong Technology Co.,Ltd. Address before: 100081 Room 204, building 3, Fuhai center, Daliushu, Haidian District, Beijing Applicant before: BEIJING SUPERRED TECHNOLOGY Co.,Ltd. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190301 |