CN106778567A - A kind of method that iris recognition is carried out by neutral net - Google Patents
A kind of method that iris recognition is carried out by neutral net Download PDFInfo
- Publication number
- CN106778567A CN106778567A CN201611102970.8A CN201611102970A CN106778567A CN 106778567 A CN106778567 A CN 106778567A CN 201611102970 A CN201611102970 A CN 201611102970A CN 106778567 A CN106778567 A CN 106778567A
- Authority
- CN
- China
- Prior art keywords
- iris
- center
- circle
- pupil
- radius
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/18—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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Ophthalmology & Optometry (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Collating Specific Patterns (AREA)
Abstract
The present invention is applied to iris recognition technology field, there is provided a kind of method for carrying out iris recognition by neutral net, including：Five neutral nets of design and training；Sample iris image is shot by infrared camera, is then found by first to fourth neutral net and is obtained the center of circle and the radius of pupil, and exterior iris boundary the center of circle and radius, finally obtain sample iris-encoding and preserve；Obtain logging in iris-encoding；To log in iris-encoding and be compared with sample iris-encoding by fifth nerve network and judge whether from same iris live body, if it is, certification passes through.Be applied to nerual network technique in iris recognition by the present invention, realize that finding the pupil center of circle and radius, the exterior iris boundary center of circle and radius, iris feature compares by designing and training five neutral nets, iris login authentication is realized, better than existing iris recognition technology.
Description
Technical field
The invention belongs to iris recognition technology field, more particularly to a kind of side that iris recognition is carried out by neutral net
Method.
Background technology
Iris recognition technology is to carry out identification based on the iris in eyes, and security device (such as gate inhibition is applied at present
Deng), and have the place of highly confidential demand.
The eye structure of people is made up of the part such as sclera, iris, pupil crystalline lens, retina.Iris is to be located at black pupil
Annular formations between hole and white sclera, it includes many interlaced spot, filament, coronal, striped, crypts etc.
Minutia.And iris is after prenatal development stage is formed, will be to maintain in whole life course constant.These are special
The uniqueness for determining iris feature is levied, while also determining the uniqueness of identification.Therefore, it can the iris of eyes is special
Levy as everyone identification object.
Current iris recognition technology is not applied to mobile terminal also, and existing iris recognition technology is directly used
Iris image parsing identification, iris recognition success rate need to be improved.
The content of the invention
In view of the above problems, it is an object of the invention to provide a kind of side that iris recognition is carried out by neutral net
Method, it is intended to solve the lower slightly technical problem of existing iris recognition technology recognition success rate.
The method of iris recognition that carried out by neutral net that the present invention is provided is applied to iris authentication system, the rainbow
Film identifying system includes general information terminal, infrared light supply, and infrared camera, the side are provided with the general information terminal
Method comprises the steps：
Five neutral nets of design and training, respectively the first to fifth nerve network, wherein first nerves network are used for
Find the pupil center of circle；Nervus opticus network is used to find the exterior iris boundary center of circle；Third nerve network be used for determine pupil radium,
Fourth nerve network is used to determine exterior iris boundary radius；Fifth nerve network is compared for iris feature；
Sample iris image is shot by infrared camera, is then found by first to fourth neutral net and is obtained pupil
The center of circle and radius, and exterior iris boundary the center of circle and radius, finally obtain sample iris-encoding and preserve；
In login authentication, shot by infrared camera and log in iris image, coordinate is set up on iris image is logged in
System, then carries out convolution using two dimension plus rich filter, obtains logging in iris-encoding；
To log in iris-encoding and be compared with sample iris-encoding by fifth nerve network and judge whether from same
One iris live body, if it is, certification passes through.
The beneficial effects of the invention are as follows：Be applied to nerual network technique in iris recognition by the present invention, by designing and instructing
Practice five neutral nets to realize that finding the pupil center of circle and radius, the exterior iris boundary center of circle and radius, iris feature compares, and realizes
Iris login authentication, experiment proves that, reach more than 98% by the inventive method disposable percent of pass of iris login authentication, it is excellent
In existing iris recognition technology.
Brief description of the drawings
Fig. 1 is the structure chart of iris authentication system；
Fig. 2 is the method flow diagram that iris recognition is carried out by neutral net；
Fig. 3 is the flow chart of step S1 in Fig. 2；
Fig. 4 is neutral net design diagram；
Fig. 5 is that neutral net five judges schematic diagram；
Fig. 6 is the control flow chart of infrared light supply energy saver mode and high effective model；
Fig. 7-1 to 7-6 is the location diagram of infrared light supply and infrared camera.
Specific embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
In order to illustrate technical solutions according to the invention, illustrated below by specific embodiment.
Fig. 1 shows a kind of structure of iris authentication system, including general information terminal, infrared light supply 2, the general letter
Infrared camera 3, including but not limited to the general information processing equipment, desktop computer, notebook are provided with breath terminal 1
Computer, smart mobile phone, panel computer, intelligence wearing, intelligent watch, intelligent glasses, Intelligent bracelet, vehicle intelligent equipment etc..Tool
Body, taken pictures offer illumination to iris by infrared light supply, the image of iris is then shot by infrared camera, find iris
The optimum position of identification.
The inventive method is realized based on above-mentioned iris authentication system, as shown in Fig. 2 comprising the steps when implementing：
Step S1, five neutral nets of design and training.
This five neutral nets are respectively the first to fifth nerve network, and wherein first nerves network is used to find pupil circle
The heart；Nervus opticus network is used to find the exterior iris boundary center of circle；Third nerve network is used to determine pupil radium, fourth nerve net
Network is used to determine exterior iris boundary radius；Fifth nerve network is compared for iris feature.
When this step is implemented, as shown in figure 3, comprising the steps：
S11, an iris image database is set up, multiple iris images of multiple users are preserved in database.
There is more than 100 user in database, each user has the iris image of more than 10.It is wherein special,
Database includes the iris image of wearing spectacles user and contact lens user, even if user's wearing spectacles can also be known
Do not go out.
S12 and then calibrate the center of circle and the radius of pupil in every iris image, and exterior iris boundary the center of circle and half
Footpath.
S13, according in the iris image database picture size design five neutral nets, as shown in figure 4, the god
Include input layer, intermediate layer and output layer through network, select input layer of the definite shape region as neutral net, input layer
Each node is mapped to the interval of [0,1] according to pixel intensity by most black value to most bright value, and the pixel in region is made
It is training sample.
In general, neutral net includes input layer, intermediate layer and output layer, and input layer is input image data, output
Layer is the result of determination of input tomographic image.Select input layer of the definite shape region as neutral net.As shown in figure 4, selected
Rectangular area, for first to fourth neutral net, when using 1920x1080 format-patterns, designs the rectangular area of 200x200
As the input layer of neutral net, it is also possible to design the rectangular area of 300x300 as the input layer of neutral net, it is also possible to make
Use other sizes.Size design may be referred to cover the maximum gauge of exterior iris boundary.The length and width of rectangular area can also be designed
Into different values.Non-rectangular area can also be used as the input layer of neutral net.Specifically, the rainbow taken out from database
Film image, with any point (x, y) as starting point, interception one size and first nerves network or nervus opticus network two it is defeated
Enter size identical rectangular area.Using this rectangle as neutral net input training sample, in order to increase the various of sample
Property, brightness or contrast can be repeatedly adjusted to same sample, it is also possible to while adjustment, it is also possible to filter using other effects
Mirror, it is also possible to use image scaled.
Each node of input layer is mapped to the area of [0,1] according to pixel intensity by most black value to most bright value
Between.
For first to fourth neutral net, it is possible to use three layers of neutral net, it is also possible to use the nerve of deeper
Network.In the neutral net using three layers, intermediate layer can preferably use 15 network nodes, it is also possible to use 20 nets
Network node, it is also possible to use the node of other quantity.
Five neutral nets specifically train as follows：
When first nerves network is trained, when the training sample center of interception is less than or equal to error with the distance in the pupil center of circle
During radius r1, training result is demarcated as 1, there is pupil in expression training sample and position is in center, when the training of interception
Center of a sample is more than error radius r1 with the distance in the pupil center of circle, or when there is no pupil in training sample, by training result mark
It is set to 0.
When nervus opticus network is trained, be less than with the distance in the exterior iris boundary center of circle when the training sample center of interception etc.
When error radius r2, training result is demarcated as 1, there is exterior iris boundary in expression training sample and position is in center.
When the training sample center of interception is more than error radius r2 with the distance in the exterior iris boundary center of circle, or there is no rainbow in training sample
During film external boundary, training result is demarcated as 0.
Because the judgement of neutral net output layer has certain uncertainty, therefore in training first nerves network and the
During two neutral nets, the span that can design error radius a r, r is greater than being equal to 0.Rule of thumb r can take 5,
10 can be taken, it is also possible to take other values.Two error radius when training the first and second neutral nets can be with identical, it is also possible to
It is different.
After the first nervus opticus network for having trained, here two neutral nets will be used to position pupil the center of circle and
The center of circle of exterior iris boundary.
In training third nerve network and fourth nerve network, during only interception is with the center of circle of pupil or exterior iris boundary
The image of the heart, in order to increase the diversity of sample, repeatedly can adjust brightness or contrast as input layer to same sample
Degree, it is also possible to while adjustment, it is also possible to use other effects filters.For the image pattern for producing radius different, it is possible to use
Image scaled.One value of radius of each node identification of output layer, during one image pattern of training, its artificial demarcation
Radius value corresponding to the training result of output node be demarcated as 1, the training result of remaining node is demarcated as 0.
For third nerve network, the nodes of output layer can be designed as the measurement of the pupil radium of iris image sample
The integer number of the pixel value of scope.Such as training sample is concentrated, and the radius span of pupil is [15,40], at this moment defeated
Go out layer and can design 26 nodes to represent 26 different results respectively.The wider array of value of more coverages can also be used
Scope.
For fourth nerve network, the nodes of its neutral net output layer can be designed as the iris of iris image sample
The integer number of the pixel value of the measurement range of external boundary.Such as training sample is concentrated, the radius value model of exterior iris boundary
It is [80,120] to enclose, and at this moment output layer can design 41 nodes and represent 41 different results respectively.Can also use more
The wider array of span of coverage.
The 3rd fourth nerve network for having trained, will be used to determine the radius of pupil and exterior iris boundary.
When fifth nerve network is trained, its output layer is a node, for representing whether two iris-encodings originate
Outside the result of determination of same iris live body, the pupil center of circle and radius, iris in a known iris image sample
On the basis of the center of circle on boundary and radius, therefrom extract iris patterns and obtain iris-encoding, each value correspondence in iris-encoding
Two iris-encodings are carried out XOR by one node of input layer, the result for obtaining as neutral net five output, such as
Really two iris-encodings derive from same iris live body, then training result is demarcated as into 1, and training result otherwise is demarcated as into 0.
Whether fifth nerve network is for judging two iris samples from same iris live body.Compiled according to iris
The size of code (iris code) designs neutral net five.The pupil center of circle in a known iris image sample and radius,
On the basis of the center of circle of exterior iris boundary and radius, therefrom extract iris patterns and obtain the mistake of iris-encoding (iris code)
Journey is prior art, and specific coding process is not repeated here.
One node of each value correspondence input layer in iris-encoding.In the neutral net using three layers, according to warp
Test, intermediate layer can preferably use 15 network nodes, it is also possible to use 20 network nodes.Other quantity can also be used
Node.
Its output layer is a node, for representing whether two iris-encodings derive from the judgement of same iris live body
As a result.When neutral net five is trained, two iris image samples are taken out from database, then according to the pupil and rainbow demarcated
Iris information is extracted out and calculates two iris-encodings by film location.Two iris-encodings are carried out into XOR, the knot for obtaining
Really as the output of neutral net five.If two iris-encodings derive from same iris live body, training result is demarcated
It is 1, training result is otherwise demarcated as 0.Whether the fifth nerve network for having trained will be used to judge two iris-encodings
From same iris live body.
Step S2, by infrared camera shoot sample iris image, then by first to fourth neutral net find
Obtain the center of circle and the radius of pupil, and exterior iris boundary the center of circle and radius, finally obtain sample iris-encoding and preserve.
After training five neutral nets, can enter iris recognition application stage, including iris information typing and
Certification.Illumination is provided to iris by infrared light supply, the image of iris is then shot by infrared camera, find iris recognition
Optimum position.
Specifically, comprising the steps：
S21, find pupil/exterior iris boundary center of circle when, one is carried out to iris image using first/second neutral net
Secondary convolution, will obtain a gray level image, and high level is will appear from around pupil, and other places are then low values；
S22, to gray level image use average filter, to remove some scattered noises；
S23 and then the pixel clearing by below threshold value n；
S24, the transverse and longitudinal coordinate finally to the pixel in high level region calculate weighted average respectively, and the result for obtaining is exactly pupil
The center of circle of hole/exterior iris boundary；
S25, after pupil/exterior iris boundary center of circle is obtained, centered on pupil/exterior iris boundary center of circle intercept a square
The image of shape, then gives three/fourth nerve network to determine the radius of pupil/exterior iris boundary；
S26, coordinate system is set up on iris image, then carry out convolution using two dimension plus rich filter, obtain sample iris
Coding.
When the center of circle of pupil is found in image pattern, a convolution is carried out to sample using first nerves network, will
To a gray level image.High level is will appear from around pupil, other places are then low values.In order to remove some scattered noises,
Can be to result images application average filter.Average filter can use 3 × 3, or 5 × 5.Other function phases can also be used
As filter.Then reuse following methods to reset the pixel of below threshold value n, it is also possible to reset using simple binaryzation.
Finally the transverse and longitudinal coordinate to the pixel in high level region calculates weighted average respectively, and the result for obtaining is exactly pupil
The center of circle.
Wherein v is the brightness value of each pixel, and x, y are respectively the transverse and longitudinal coordinates of each pixel.
, it is necessary to the region of predefined eyes appearance, then individually calculates when left and right eyes are recognized.
It is same as mentioned above when the center of circle of exterior iris boundary is found in image pattern, sought using nervus opticus network
Look for the center of circle of exterior iris boundary.
An image for rectangle is intercepted centered on the pupil center of circle, then gives third nerve network to determine the half of pupil
Footpath.An image for rectangle is intercepted centered on the exterior iris boundary center of circle, then gives fourth nerve network to determine outside iris
The radius on border.
After finding iris inner and outer boundary, coordinate system is set up on iris image, then carried out using two dimension plus rich filter
Convolution, obtains sample iris-encoding.In iris information typing, by sample iris-encoding encrypting storing to locally.
Step S3, in login authentication, by infrared camera shoot log in iris image, log in iris image on build
Vertical coordinate system, then carries out convolution using two dimension plus rich filter, obtains logging in iris-encoding；
Step S4, will be logged in by fifth nerve network iris-encoding and sample iris-encoding be compared judge whether come
Same iris live body is come from, if it is, certification passes through.
In the login authentication stage, adopted when the sample iris-encoding of advance typing is judged using fifth nerve network with checking
Whether the login iris-encoding obtained by the iris image of collection derives from same iris live body.Two iris-encoding step-by-steps are carried out
XOR, and fifth nerve network is given by result, obtain result of determination.Specifically, logging in rainbow as shown in figure 5, first calculating
Film encodes the Hamming distances with sample iris-encoding：When Hamming distances are less than threshold value a, directly it is judged to from same rainbow
Film live body；When Hamming distances are more than threshold value b, directly it is judged to from different iris live bodies；At Hamming distances with a with
When between b, will log in iris-encoding carries out XOR with the step-by-step of sample iris-encoding, and operation result is used as fifth nerve network
Input layer, be then judged to from same iris live body when fifth nerve network is output as 1, by certification.
Present invention method designs five neutral nets, wherein first to fourth neutral net is used to realize finding
The pupil center of circle and radius, the exterior iris boundary center of circle and radius, fifth nerve network are used to realize that iris feature is compared and login is recognized
Card, experiment proves that, more than 98% is reached by the inventive method disposable percent of pass of iris login authentication, better than existing iris
Identification technology.
Can also be that the infrared light supply sets the energy saver mode that an interval long is flashed additionally as preferred embodiment
It is short with one to be spaced the high effective model for flashing, when iris image is searched out, into high effective model, stepped in no iris for a long time
During record acts of authentication, into energy saver mode.Specific control flow as shown in fig. 6, first turn on the infrared camera of iris recognition,
Infrared light supply is opened, into the energy saver mode that interval long is flashed, then people's eye iris image is gathered and is passed through neural net method
Iris position is found, if finding iris position, high effective model is flashed into short interval, then iris image is carried out to have processed
Into logining or identifying procedure；If can not find iris position, when apart from last time found the iris time more than threshold value when, be spaced into long
Energy saver mode is flashed, otherwise continues to gather eye iris image.Because the energy content of battery of general information terminal is limited, this programme can
It is embodied as general information terminal and intelligent management is provided.Flashing time interval and should take into account iris illumination equipment in high effective model
Temperature is tolerated and service life.During using multiple infrared light supplies, the phase flashed can be adjusted to obtain more preferably more continuous photograph
Obvious results is really.
Because the energy content of battery of general information terminal is limited, a lens or saturating can be placed in the front equipment end of infrared illumination
Microscope group (based on convex lens, comprising Fresnel Lenses) improves light utilization, infrared light supply and convex lens with convergent light rays
Distance is necessarily less than the focal length of convex lens to ensure the security of light.Convergent divergence of beam angle should be less than 45 degree.It is excellent
Choosing, dispersion angle should be less than 30 degree.Infrared camera using the module that focuses of 20cm-40cm, or can be used same
The autozoom module of scope.
Iris shoot optimum position is in the front of infrared camera and watches infrared camera attentively, infrared light supply is most
Best placement should be close proximity to infrared camera.In order to help user to find iris recognition optimum position, in general information terminal
Screen on show the image of user, the image that at this moment use habit of user can be watched attentively in screen, without watching infrared taking the photograph attentively
As head, thus should by infrared camera design screen top centre or lower section centre, close proximity to display
Image.If cannot by infrared camera design in screen either above or below centre, the image of shooting will it is to the left or
Person is to the right, it is preferred that can be translated by software mode or interception image and then display.Can also be without translating or cutting
Take.When being illuminated using an infrared light supply, should be close proximity to infrared camera.Illuminated using the infrared light supply of two and the above
When, should be distributed in around infrared camera.In order to avoid the image that infrared light supply direct interference camera shoots, both it
Between distance should be maintained at more than 1mm.Or increase ensures that infrared light supply will not pass through camera antetheca every photosphere during design
Direct interference image.General information terminal realizes structure for smart mobile phone is that one kind is most common, and Fig. 7-1 to Fig. 7-6 shows respectively
Several positions of infrared light supply and infrared camera are gone out to show, in Fig. 7-1 to 7-3, infrared light supply and infrared camera exist
Screen top, in Fig. 7-4 to 7-6, infrared light supply and infrared camera below screen, in Fig. 7-3 and Fig. 7-6, infrared light supply
There are two, centre is infrared camera.The present embodiment includes but is not limited to this six kinds of position relationships.
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the invention, it is all in essence of the invention
Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.
Claims (6)
1. a kind of method for carrying out iris recognition by neutral net, the method is applied to iris authentication system, the iris
Identifying system includes general information terminal, infrared light supply, and infrared camera, methods described are provided with the general information terminal
Comprise the steps：
Five neutral nets of design and training, respectively the first to fifth nerve network, wherein first nerves network are used to find
The pupil center of circle；Nervus opticus network is used to find the exterior iris boundary center of circle；Third nerve network is used to determine pupil radium, the 4th
Neutral net is used to determine exterior iris boundary radius；Fifth nerve network is compared for iris feature；
Sample iris image is shot by infrared camera, the circle for obtaining pupil is then found by first to fourth neutral net
The heart and radius, and exterior iris boundary the center of circle and radius, finally obtain sample iris-encoding and preserve；
In login authentication, shot by infrared camera and log in iris image, set up coordinate system on iris image is logged in, so
Convolution is carried out using two dimension plus rich filter afterwards, obtains logging in iris-encoding；
To log in iris-encoding and be compared with sample iris-encoding by fifth nerve network and judge whether from same
Iris live body, if it is, certification passes through.
2. method as claimed in claim 1, it is characterised in that five neutral net steps of the design and training, specifically includes：
An iris image database is set up, multiple iris images of multiple users are preserved in database.
Then calibrate the center of circle and the radius of pupil in every iris image, and exterior iris boundary the center of circle and radius.
According in the iris image database picture size design five neutral nets, the neutral net include input layer,
Intermediate layer and output layer, select input layer of the definite shape region as neutral net, and each node of input layer is according to pixel
Brightness is mapped to the interval of [0,1] by most black value to most bright value, the pixel in region as training sample, five god
Specifically trained through network as follows：
When first nerves network is trained, when the training sample center of interception is less than or equal to error radius with the distance in the pupil center of circle
During r1, training result is demarcated as 1, there is pupil in expression training sample and position is in center, when the training sample of interception
Center is more than error radius r1 with the distance in the pupil center of circle, or when not having pupil in training sample, training result is demarcated as
0；
When nervus opticus network is trained, when the training sample center of interception with the distance in the exterior iris boundary center of circle less than or equal to mistake
During difference radius r2, training result is demarcated as 1, there is exterior iris boundary in expression training sample and position is in center.When cut
The training sample center for taking is more than error radius r2 with the distance in the exterior iris boundary center of circle, or does not have outside iris in training sample
During border, training result is demarcated as 0；
In training third nerve network and fourth nerve network, only interception is centered on the center of circle of pupil or exterior iris boundary
Image as input layer, one value of radius of each node identification of output layer, during one image pattern of training, its artificial mark
The training result of the output node corresponding to fixed radius value is demarcated as 1, and the training result of remaining node is demarcated as 0；
Whether, when fifth nerve network is trained, its output layer is a node, for representing two iris-encodings from same
One result of determination of iris live body, the pupil center of circle in a known iris image sample and radius, exterior iris boundary
On the basis of the center of circle and radius, therefrom extract iris patterns and obtain iris-encoding, each value correspondence in iris-encoding is input into
Layer a node, two iris-encodings are carried out into XOR, the result for obtaining as neutral net five output, if two
Individual iris-encoding derives from same iris live body, then training result is demarcated as into 1, and training result otherwise is demarcated as into 0.
3. method as claimed in claim 2, it is characterised in that described to pass through first to fourth neutral net and find to obtain pupil
The center of circle and radius, and exterior iris boundary the center of circle and radius, finally obtain sample iris-encoding and preserve step, specific bag
Include：
When pupil/exterior iris boundary center of circle is found, a convolution is carried out to iris image using first/second neutral net,
A gray level image will be obtained, high level will be will appear from around pupil, other places will be then low values；
Average filter is used to gray level image, to remove some scattered noises；
Then the pixel of below threshold value n is reset；
Finally the transverse and longitudinal coordinate to the pixel in high level region calculates weighted average respectively, and the result for obtaining is exactly pupil/iris
The center of circle of external boundary；
After pupil/exterior iris boundary center of circle is obtained, a figure for rectangle is intercepted centered on pupil/exterior iris boundary center of circle
Picture, then gives three/fourth nerve network to determine the radius of pupil/exterior iris boundary；
Coordinate system is set up on iris image, then convolution is carried out using two dimension plus rich filter, sample iris-encoding is obtained.
4. method as claimed in claim 3, it is characterised in that described that iris-encoding and sample will be logged in by fifth nerve network
Iris-encoding is compared and judges whether from same iris live body, if it is, certification is specifically included by step：
First calculate the Hamming distances for logging in iris-encoding and sample iris-encoding；
When Hamming distances are less than threshold value a, directly it is judged to from same iris live body；
When Hamming distances are more than threshold value b, directly it is judged to from different iris live bodies；
When at Hamming distances and a and b between, will log in iris-encoding carries out XOR with the step-by-step of sample iris-encoding, transports
Input layer of the result as fifth nerve network is calculated, is then judged to from same when fifth nerve network is output as 1
Individual iris live body, by certification.
5. such as claim 1-4 any one methods describeds, it is characterised in that the infrared light supply has the section that an interval long is flashed
The high effective model that energy pattern and a short interval are flashed, when iris image is searched out, into high effective model, does not have in long-time
When iris authentication is acted, into energy saver mode.
6. method as claimed in claim 5, it is characterised in that the infrared light supply front end be placed with a lens or lens group with
Convergent light rays improve light utilization, and the infrared light supply is necessarily less than the focal length of convex lens to ensure light with the distance of convex lens
The security of line, convergent divergence of beam angle is less than 45 degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611102970.8A CN106778567B (en) | 2016-12-05 | 2016-12-05 | A method of iris recognition is carried out by neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611102970.8A CN106778567B (en) | 2016-12-05 | 2016-12-05 | A method of iris recognition is carried out by neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106778567A true CN106778567A (en) | 2017-05-31 |
| CN106778567B CN106778567B (en) | 2019-05-28 |
Family
ID=58883813
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611102970.8A Active CN106778567B (en) | 2016-12-05 | 2016-12-05 | A method of iris recognition is carried out by neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106778567B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480659A (en) * | 2017-09-21 | 2017-12-15 | 深圳普创天信科技发展有限公司 | A kind of iris identification device and terminal |
| CN107506721A (en) * | 2017-08-18 | 2017-12-22 | 上海爱优威软件开发有限公司 | A kind of more method for collecting iris and system |
| CN107516080A (en) * | 2017-08-22 | 2017-12-26 | 京东方科技集团股份有限公司 | The Ins location method, apparatus and electronic equipment of PVR image |
| WO2018222897A1 (en) * | 2017-06-01 | 2018-12-06 | University Of Washington | Smartphone-based digital pupillometer |
| CN109165538A (en) * | 2018-07-18 | 2019-01-08 | 北京飞搜科技有限公司 | Bar code detection method and device 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 |
| CN111031241A (en) * | 2019-12-09 | 2020-04-17 | Oppo广东移动通信有限公司 | Image processing method and device, terminal and computer readable storage medium |
| CN111278351A (en) * | 2017-07-28 | 2020-06-12 | 瞳孔扫描公司 | System and method for capturing and analyzing pupil images to determine toxicology and neurophysiology |
| CN111488843A (en) * | 2020-04-16 | 2020-08-04 | 贵州安防工程技术研究中心有限公司 | Face sunglasses distinguishing method based on step-by-step inhibition of missing report and false report rate |
| CN113553908A (en) * | 2021-06-23 | 2021-10-26 | 中国科学院自动化研究所 | Heterogeneous iris identification method based on equipment unique perception |
| WO2021216096A1 (en) * | 2020-04-19 | 2021-10-28 | Pupilscan Corporation | Self-scanning ophthalmic device |
| CN114092701A (en) * | 2021-12-04 | 2022-02-25 | 特斯联科技集团有限公司 | Intelligent symbol identification method based on neural network |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2006025129A1 (en) * | 2004-08-30 | 2006-03-09 | Toyama-Prefecture | Personal authentication system |
| CN102306289A (en) * | 2011-09-16 | 2012-01-04 | 兰州大学 | Method for extracting iris features based on pulse couple neural network (PCNN) |
| CN103246871A (en) * | 2013-04-25 | 2013-08-14 | 山东师范大学 | Non-ideal iris outer boundary positioning method based on image nonlinear enhancement |
| CN104751186A (en) * | 2015-04-10 | 2015-07-01 | 山东师范大学 | Iris image quality classification method based on BP (back propagation) network and wavelet transformation |
| CN105303185A (en) * | 2015-11-27 | 2016-02-03 | 中国科学院深圳先进技术研究院 | Iris positioning method and device |
| CN105488487A (en) * | 2015-12-09 | 2016-04-13 | 湖北润宏科技有限公司 | Iris positioning method and device |
-
2016
- 2016-12-05 CN CN201611102970.8A patent/CN106778567B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2006025129A1 (en) * | 2004-08-30 | 2006-03-09 | Toyama-Prefecture | Personal authentication system |
| CN102306289A (en) * | 2011-09-16 | 2012-01-04 | 兰州大学 | Method for extracting iris features based on pulse couple neural network (PCNN) |
| CN103246871A (en) * | 2013-04-25 | 2013-08-14 | 山东师范大学 | Non-ideal iris outer boundary positioning method based on image nonlinear enhancement |
| CN104751186A (en) * | 2015-04-10 | 2015-07-01 | 山东师范大学 | Iris image quality classification method based on BP (back propagation) network and wavelet transformation |
| CN105303185A (en) * | 2015-11-27 | 2016-02-03 | 中国科学院深圳先进技术研究院 | Iris positioning method and device |
| CN105488487A (en) * | 2015-12-09 | 2016-04-13 | 湖北润宏科技有限公司 | Iris positioning method and device |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200129063A1 (en) | 2017-06-01 | 2020-04-30 | University Of Washington | Smartphone-based digital pupillometer |
| WO2018222897A1 (en) * | 2017-06-01 | 2018-12-06 | University Of Washington | Smartphone-based digital pupillometer |
| US20230225612A1 (en) * | 2017-06-01 | 2023-07-20 | University Of Washington | Smartphone-based digital pupillometer |
| US11647903B2 (en) | 2017-06-01 | 2023-05-16 | University Of Washington | Smartphone-based digital pupillometer |
| CN111278351A (en) * | 2017-07-28 | 2020-06-12 | 瞳孔扫描公司 | System and method for capturing and analyzing pupil images to determine toxicology and neurophysiology |
| CN107506721A (en) * | 2017-08-18 | 2017-12-22 | 上海爱优威软件开发有限公司 | A kind of more method for collecting iris and system |
| CN107516080A (en) * | 2017-08-22 | 2017-12-26 | 京东方科技集团股份有限公司 | The Ins location method, apparatus and electronic equipment of PVR image |
| US10772501B2 (en) | 2017-08-22 | 2020-09-15 | Boe Technology Group Co., Ltd. | Method and electronic device for positioning eyeball in retinopathy image |
| WO2019056623A1 (en) * | 2017-09-21 | 2019-03-28 | 深圳普创天信科技发展有限公司 | Iris recognition device and terminal |
| CN107480659A (en) * | 2017-09-21 | 2017-12-15 | 深圳普创天信科技发展有限公司 | A kind of iris identification device and terminal |
| CN109165538B (en) * | 2018-07-18 | 2021-10-22 | 苏州飞搜科技有限公司 | Bar code detection method and device based on deep neural network |
| CN109165538A (en) * | 2018-07-18 | 2019-01-08 | 北京飞搜科技有限公司 | Bar code detection method and device based on deep neural network |
| CN110046588B (en) * | 2019-04-22 | 2019-11-01 | 吉林大学 | It is a kind of with steal attack coping mechanism heterogeneous iris recognition method |
| 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 |
| CN111031241A (en) * | 2019-12-09 | 2020-04-17 | Oppo广东移动通信有限公司 | Image processing method and device, terminal and computer readable storage medium |
| CN111031241B (en) * | 2019-12-09 | 2021-08-27 | Oppo广东移动通信有限公司 | Image processing method and device, terminal and computer readable storage medium |
| CN111488843A (en) * | 2020-04-16 | 2020-08-04 | 贵州安防工程技术研究中心有限公司 | Face sunglasses distinguishing method based on step-by-step inhibition of missing report and false report rate |
| WO2021216096A1 (en) * | 2020-04-19 | 2021-10-28 | Pupilscan Corporation | Self-scanning ophthalmic device |
| CN113553908A (en) * | 2021-06-23 | 2021-10-26 | 中国科学院自动化研究所 | Heterogeneous iris identification method based on equipment unique perception |
| CN114092701A (en) * | 2021-12-04 | 2022-02-25 | 特斯联科技集团有限公司 | Intelligent symbol identification method based on neural network |
| CN114092701B (en) * | 2021-12-04 | 2022-06-03 | 特斯联科技集团有限公司 | Intelligent symbol identification method based on neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106778567B (en) | 2019-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106778567B (en) | A method of iris recognition is carried out by neural network | |
| CN106778664B (en) | Iris image iris area segmentation method and device | |
| CN112651348B (en) | Identity authentication method and device and storage medium | |
| CN103106401B (en) | Mobile terminal iris recognition device with human-computer interaction mechanism | |
| Sharif et al. | Accessorize to a crime: Real and stealthy attacks on state-of-the-art face recognition | |
| CN108427503B (en) | Human eye tracking method and human eye tracking device | |
| JP3940759B2 (en) | Iris registration method, iris registration device, and iris registration program | |
| US9798927B2 (en) | Mobile terminal iris recognition method and device having human-computer interaction mechanism | |
| CN103383723B (en) | Method and system for spoof detection for biometric authentication | |
| US10445574B2 (en) | Method and apparatus for iris recognition | |
| KR100954640B1 (en) | Personal authentication method and device | |
| CN204129753U (en) | iris identification device | |
| CN110210276A (en) | A kind of motion track acquisition methods and its equipment, storage medium, terminal | |
| CN105046250B (en) | The glasses removing method of recognition of face | |
| CN108898171A (en) | Recognition processing method, system and computer readable storage medium | |
| CN110287900A (en) | Verification method and verifying device | |
| Li et al. | Light can hack your face! black-box backdoor attack on face recognition systems | |
| CN111353404A (en) | Face recognition method, device and equipment | |
| CN109255282B (en) | Biological identification method, device and system | |
| US20230386429A1 (en) | Systems and Methods for Switching Vision Correction Graphical Outputs on a Display of an Electronic Device | |
| CN115937959A (en) | Method and device for determining gazing information and eye movement tracking equipment | |
| CN109697735A (en) | Visible light image processing method for positioning in faced chamber under intelligent lamp | |
| CN105631285A (en) | Biological feature identity recognition method and apparatus | |
| CN107220623A (en) | A kind of face identification method and system | |
| CN113920591A (en) | Medium-distance and long-distance identity authentication method and device based on multi-mode biological feature recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |