CN109815850A - Iris segmentation and localization method, system, device based on deep learning - Google Patents
Iris segmentation and localization method, system, device based on deep learning Download PDFInfo
- Publication number
- CN109815850A CN109815850A CN201910002663.XA CN201910002663A CN109815850A CN 109815850 A CN109815850 A CN 109815850A CN 201910002663 A CN201910002663 A CN 201910002663A CN 109815850 A CN109815850 A CN 109815850A
- Authority
- CN
- China
- Prior art keywords
- iris
- mapping graph
- segmentation
- boundary
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 134
- 238000000034 method Methods 0.000 title claims abstract description 76
- 238000013135 deep learning Methods 0.000 title claims abstract description 48
- 230000004807 localization Effects 0.000 title claims abstract description 34
- 210000000554 iris Anatomy 0.000 claims abstract description 315
- 238000013507 mapping Methods 0.000 claims abstract description 144
- 210000001747 pupil Anatomy 0.000 claims abstract description 31
- 238000003062 neural network model Methods 0.000 claims abstract description 16
- 210000003786 sclera Anatomy 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 9
- 230000000644 propagated effect Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 230000000149 penetrating effect Effects 0.000 claims 2
- 238000010606 normalization Methods 0.000 abstract description 4
- 238000003909 pattern recognition Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 101100295091 Arabidopsis thaliana NUDT14 gene Proteins 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 230000033001 locomotion Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 210000000720 eyelash Anatomy 0.000 description 3
- 208000004350 Strabismus Diseases 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 210000004209 hair Anatomy 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 210000000744 eyelid Anatomy 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 210000003128 head Anatomy 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Abstract
The invention belongs to pattern-recognition, computer vision and field of image processings, and in particular to a kind of iris segmentation and localization method, system, device based on deep learning, it is intended to solve the problems, such as that iris recognition precision is not high under non-controllable scene.The method of the present invention includes: to obtain iris image to be processed；Four mapping graphs are generated using multitask neural network model, respectively correspond pupil center, iris inner boundary, exterior iris boundary and iris segmentation mask；Iris segmentation mask mapping graph is handled using Threshold segmentation, completes iris segmentation；Pupil center location is predicted according to the geometrical relationship of pupil center and iris mask；Mapping graph is denoised and calculated using the geometrical relationship between pupil, iris, sclera, Circle Parameters inside and outside iris is obtained, completes Iris Location.The present invention can carry out effective segmentation positioning to the iris image acquired under non-controllable environment, lay a good foundation for subsequent normalization and identification, improve the precision of iris recognition under non-controllable environment.
Description
Technical field
The invention belongs to pattern-recognition, computer vision and field of image processing, and in particular to a kind of to be based on depth
Iris segmentation and localization method, system, the device of habit.
Background technique
With the rise of artificial intelligence, the biometrics identification technologies such as recognition of face, iris recognition, fingerprint recognition are received
Greatly concern, wherein iris recognition technology is considered as one of most stable, most accurate and most reliable verification method, therefore wide
It is general to be applied to the fields such as intelligent unlocking, border control, bank finance, access control and attendance.
In iris authentication system, the start-up portion of iris segmentation and positioning in entire process flow, therefore it is accurate
Property will directly affect the precision of subsequent processing.Iris segmentation, which refers to, extracts effective iris texture pixel, excludes noise, final output
The segmentation mask of two-value.Iris Location refers to the inside and outside circle boundary parameter that iris is accurately positioned.The result of Iris Location is used in
The normalization operation of iris；And the mask segmentation of iris will be related to the processing of iris image noise region.
Traditional iris segmentation localization method is usually to be combined, and may be collectively referred to as iris segmentation.Typical segmentation
Method can be divided into two major classes: first is that the method based on edge, need to position respectively the outer edge of iris, upper lower eyelid and
Removal eyelashes block etc. to obtain isolated iris region；Second is that method pixel-based, directly according to outer near pixel
Feature, such as color, texture etc. are seen, directly difference iris pixel and non-iris pixel.Generally speaking, these methods usually require
A large amount of priori knowledge and many intermediary operations are relied on, process is complicated, is often suitable only for containing clear iris inner and outer boundary
Iris image.With a wide range of universal, the iris image of acquisition of distant range iris identification and mobile terminal iris recognition etc.
Often due to illumination, target movement and distance change, cause containing mirror-reflection, strabismus, obscure, glasses such as block at various make an uproar
Sound, traditional method cannot handle this kind of image well.
Therefore, in order to effectively carry out accurate pretreatment operation to the iris image acquired under non-controllable scene,
There is an urgent need to develop a kind of new, accurate, efficient iris segmentation and localization methods, meet the requirement of user, effectively
Improve the precision of the iris recognition under non-controllable scene.
Summary of the invention
In order to solve the above problem in the prior art, i.e., the not high problem of iris recognition precision under non-controllable scene, this
Invention provides a kind of iris segmentation method based on deep learning, comprising:
Step S10 obtains iris image to be processed, as the first iris image；
First iris image is inputted trained full convolution encoding and decoding multitask neural network model by step S20
Propagated forward is carried out, first pupil center's mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph are obtained
With the first iris segmentation mask mapping graph；
Step S30 handles the first iris segmentation mask mapping graph using Threshold segmentation, obtains two-value iris segmentation and covers
Mould figure completes iris segmentation；
The trained full convolution encoding and decoding multitask neural network model, acquisition modes are as follows:
Step S21 obtains iris image and second iris image is marked as the second iris image, obtains
Third iris image；Attention mechanism is incorporated, the encoding and decoding multitask neural network of full convolution is constructed；
Step S22 instructs the encoding and decoding multitask neural network that the third iris image inputs the full convolution
Practice, obtains the encoding and decoding multitask neural network model of trained full convolution.
In some preferred embodiments, step S21 " is marked the second iris image ", the steps include:
Step S211 is marked iris pixel effective in second iris image using binary code label；
Effective iris pixel is labeled as 1, remaining position is labeled as 0；Using two ellipses closest to iris inner and outer boundary, as rainbow
The label of the inner and outer boundary of film；Boundary point is labeled as 1, remaining position is labeled as 0；
The effective iris pixel, without non-irises such as noises and hair, eyelashes, pupil, sclera such as mirror-reflections
Region.
Step S212, the centre mark by boundary oval in the iris marked are the center of pupil, and it includes effective for obtaining
The iris image of iris separator, iris inner and outer boundary label and pupil center's label is as third iris image；The rainbow
Pupil center's point of film image is labeled as 1, remaining position is labeled as 0.
In some preferred embodiments, the encoding and decoding multitask neural network of the full convolution includes coding layer and decoding
Layer.
The coding layer is operated by multiple continuous convolution, ReLu and pondization come coding characteristic, to being originally inputted figure
The mapping graph resolution ratio of picture is continuously reduced, and low resolution mapping graph is obtained.
The decoding layer carries out jump connection by layer corresponding with decoding layer, is up-sampled using bilinearity to the coding
Low resolution mapping graph after the continuous diminution of layer carries out resolution ratio recovery, exports four mappings identical with size is originally inputted
Figure, corresponding pupil center's mapping graph, iris inner boundary mapping graph, exterior iris boundary mapping graph and iris segmentation mask mapping graph.
In some preferred embodiments, " the third iris image is inputted into the full convolution described in step S22
Encoding and decoding multitask neural network is trained ", it the steps include:
Step S221 inputs the third iris image in the encoding and decoding multitask neural network of the full convolution, leads to
Cross propagated forward obtain second pupil center's mapping graph, the second iris inner boundary mapping graph, the second exterior iris boundary mapping graph and
Second iris segmentation mask mapping graph.
Step S222 calculates second pupil center mapping graph, the second iris inner boundary mapping graph, the second iris outside
The error of boundary's mapping graph and the second iris segmentation mask mapping graph and the third iris image, based on the resulting total damage of calculating
Functional value is lost, the encoding and decoding multitask neural network of full convolution is joined using back-propagation algorithm and stochastic gradient descent method
Number updates.
Step S223 repeats step S222, until total loss function calculated value reaches preset condition, is trained
Full convolution encoding and decoding multitask neural network model.
In some preferred embodiments, total loss function includes the focal loss for being directed to pupil center, is directed to
The intersection entropy loss of the balance of iris inner and outer boundary and the intersection entropy loss for segmentation；
The focal loss L for pupil centerpupilAre as follows:
Wherein, alpha, gamma is hyper parameter；P={ pj, j=1 ..., | X | }, pjIt is for j-th of pixel of multitask neural network forecast
The probability of pupil center, | X | it is the number of pixels of iris image； For jth in iris image
The label of a pixel, 1 is expressed as true iris pupil center, and 0 is not to be.
The intersection entropy loss L of the balance for iris inner and outer boundaryedgeAre as follows:
Wherein, Indicate that j-th of pixel belongs to iris inner boundary in iris image
Or the probability of exterior iris boundary, k=1 represent iris inner boundary, k=2 represents exterior iris boundary, and 1 is expressed as true iris
Boundary, 0 is not to be；β is ratio shared by non-edge pixels in iris image, and non-edge pixels includes iris inner boundary and outside
Boundary's pixel.
The intersection entropy loss for segmentation is LsegAre as follows:
Wherein, S={ sj, j=1 ... | X |, sjIndicate that j-th of pixel belongs to true iris texture in iris image
Probability； J-th of pixel belongs to the label of true iris texture in expression iris image.
Another aspect of the present invention proposes a kind of iris image localization method based on deep learning, comprising:
Using the step S10- step S20 of the above-mentioned iris segmentation method based on deep learning, the first pupil is obtained
Centralizing mapping figure, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and the first iris segmentation mask mapping graph,
And execute following steps:
Step B10, according to the geometry of first pupil center mapping graph and the first iris segmentation mask mapping graph
Relationship obtains the predicted position of pupil center；
Step B20, based on the geometrical relationship between preset pupil, iris, sclera, to the first iris inner and outer boundary
Mapping graph denoising, obtains circle center origin coordinates inside and outside the first iris inside and outside circle radius parameter collection and the first iris；
Step B30 is joined based on radius of circle inside and outside the first iris inner and outer boundary mapping graph and first iris after denoising
Circle center origin coordinates inside and outside manifold and the first iris obtains the match point of iris inside and outside contour using Viterbi algorithm, uses
The circle fitting algorithm of least square is fitted inside and outside circle, obtains Circle Parameters inside and outside iris, completes Iris Location.
In some preferred embodiments, " according to first pupil center mapping graph and first rainbow in step B10
The geometrical relationship of film segmentation mask mapping graph, obtains the predicted position of pupil center ", method are as follows:
Threshold segmentation and connected region are carried out to first pupil center mapping graph and the first iris segmentation mask mapping graph
Communicated subarea of the communicated subarea of domain analysis, search and maximum area segmentation mask apart from nearest pupil center, takes institute
Stating communicated subarea central point is the pupil center finally predicted.
In some preferred embodiments, " closed based on the geometry between preset pupil, iris, sclera in step B20
System denoises the first iris inner and outer boundary mapping graph, obtains in the first iris inside and outside circle radius parameter collection and the first iris
Outer circle center origin coordinates ", the steps include:
Step B21, it is maximum to the dicing masks by the center of circle, the pupil center of the predicted position of the pupil center
The maximum distance of the communicated subarea of area is radius, generates limited area；
Step B22 deletes the inside and outside exterior iris boundary mapping graph connected region of the limited area, obtains iris
Outer boundary；The iris inner boundary mapping graph connected region outside the limited area is deleted, iris inner boundary is obtained；
Step B23, the minimax distance of the iris inner and outer boundary of pupil center to the refinement is as in the first iris
Exradius parameter set, center coordinate of eye pupil are circle center origin coordinates inside and outside the first iris.
The third aspect of the present invention proposes a kind of iris segmentation and positioning system based on deep learning, including
Input module, multitask neural network module, Threshold segmentation module, pupil position prediction module, denoising module, fitting module,
Output module；
The input module is configured to obtain iris image to be tested, as the first iris image；
The multitask neural network module is configured to obtain the first pupil center to first iris image processing
Mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and the first iris segmentation mask mapping graph；
The Threshold segmentation module is configured to handle the first iris segmentation mask mapping graph using Threshold segmentation, obtain
Two-value iris segmentation mask artwork is obtained, iris segmentation is completed；
The pupil position prediction module is configured to be covered according to first pupil center mapping graph and the first iris segmentation
The geometrical relationship of Module map figure obtains the predicted position of pupil center；
The denoising module is configured to preset pupil, iris, the geometrical relationship between sclera to first rainbow
The denoising of film inner and outer boundary mapping graph obtains circle center starting inside and outside the first iris inside and outside circle radius parameter collection and the first iris and sits
Mark；
The fitting module is configured in the first iris inner and outer boundary mapping graph and first iris after denoising
Circle center origin coordinates inside and outside exradius parameter set and the first iris obtains iris inside and outside contour using Viterbi algorithm
Match point is fitted inside and outside circle using the circle fitting algorithm of least square, obtains Circle Parameters inside and outside iris, complete Iris Location；
The output module is configured to export Circle Parameters inside and outside the two-value iris segmentation mask artwork and iris.
The fourth aspect of the present invention proposes a kind of storage device, wherein be stored with a plurality of program, described program be suitable for by
Processor is loaded and is executed to realize above-mentioned iris segmentation and localization method based on deep learning.
The fifth aspect of the present invention proposes a kind of processing unit, including processor, storage device；The processor is fitted
In each program of execution；The storage device is suitable for storing a plurality of program；Described program be suitable for loaded by processor and executed with
Realize above-mentioned iris segmentation and localization method based on deep learning.
Beneficial effects of the present invention:
(1) compared with prior art compared with, can be to being acquired under non-controllable scene the present invention is based on deep learning frame
Iris image carries out accurate, robust Ground Split and positioning, meets the requirement of user, and can improve the accuracy of iris recognition, has
There is great production practices meaning.
(2) the present invention is based on deep learning frame, the multitask network of proposition effectively captures multi-modal eye structure
Geometry associativity, while propose attention mechanism and DeepLab ASPP module or PSPNet pyramid pond module
It combines, network is allowed to capture the feature of optimal judgement index, and the accurate segmentation mask of final output and other mode
Structure.The post-processing operation of subsequent proposition can effectively eliminate the interference of noise, be accurately located the inside and outside circle of iris.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is the flow chart the present invention is based on the iris segmentation of deep learning and localization method；
Fig. 2 is that the present invention is based on adopt under the non-controllable scene of the iris segmentation of deep learning and localization method embodiment
The schematic diagram of the iris image of collection；
Fig. 3 is the multitask network structure the present invention is based on the iris segmentation of deep learning and localization method embodiment
Schematic diagram；
Fig. 4 is the integrated attention mechanism the present invention is based on the iris segmentation of deep learning and localization method embodiment
With the ASPP modular structure schematic diagram of Deeplab；
Fig. 5 is the integrated attention mechanism the present invention is based on the iris segmentation of deep learning and localization method embodiment
With PSPNet pyramid pond modular structure schematic diagram；
Fig. 6 is that the present invention is based on the iris segmentations of deep learning and the multitask network of localization method embodiment to export
Four mapping graphs and the contrast schematic diagram with former iris image；
Fig. 7 be the present invention is based on the iris segmentation of deep learning and localization method embodiment according to pupil center with
The flow chart of geometrical relationship prediction pupil center's point of iris mask；
Fig. 8 is the inner and outer boundary mapping graph the present invention is based on the iris segmentation of deep learning and localization method embodiment
Denoising obtains the flow chart of the range parameter of iris inside and outside circle；
Fig. 9 is that the present invention is based on the iris segmentation of deep learning and the iris masks of localization method embodiment, inside and outside
The round contrast schematic diagram with former iris image.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is only used for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to just
Part relevant to related invention is illustrated only in description, attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
The present invention provides a kind of iris segmentation and localization method based on deep learning, and this method contains two
Point: first part is the coding layer decoding layer multitask network based on full convolution, and the coding layer of network is integrated with attention
The pyramid pond layer of the ASPP module or PSPNet of mechanism and DeepLab, the mask that this stage can export segmentation are completed
Segmentation task；Second part is an effective post-processing approach, for carrying out the positioning of iris.This method is applied to low-quality
When iris image, noise jamming can be effectively eliminated, the mask of segmentation can be provided, the parameter of inside and outside circle can be also provided, be
Subsequent iris recognition process is laid a good foundation.
A kind of iris segmentation method based on deep learning of the invention, comprising:
Step S10 obtains iris image to be processed, as the first iris image；
First iris image is inputted trained full convolution encoding and decoding multitask neural network model by step S20
Propagated forward is carried out, first pupil center's mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph are obtained
With the first iris segmentation mask mapping graph；
Step S30 handles the first iris segmentation mask mapping graph using Threshold segmentation, obtains two-value iris segmentation and covers
Mould figure completes iris segmentation；
The trained full convolution encoding and decoding multitask neural network model, acquisition modes are as follows:
Step S21 obtains iris image and second iris image is marked as the second iris image, obtains
Third iris image；Attention mechanism is incorporated, the encoding and decoding multitask neural network of full convolution is constructed；
Step S22 instructs the encoding and decoding multitask neural network that the third iris image inputs the full convolution
Practice, obtains the encoding and decoding multitask neural network model of trained full convolution.
In order to be more clearly illustrated to the iris segmentation method the present invention is based on deep learning, below with reference to figure
Each step expansion is described in detail in 1 pair of embodiment of the present invention method.
The iris image localization method based on deep learning of an embodiment of the present invention, including step S10- step S30,
Each step is described in detail as follows:
Step S10 obtains iris image to be processed, as the first iris image.
First iris image is inputted trained full convolution encoding and decoding multitask neural network model by step S20
Propagated forward is carried out, first pupil center's mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph are obtained
With the first iris segmentation mask mapping graph.
The trained full convolution encoding and decoding multitask neural network model, acquisition modes are as follows:
Step S21 obtains iris image and second iris image is marked as the second iris image, obtains
Third iris image；Attention mechanism is incorporated, the encoding and decoding multitask neural network of full convolution is constructed.
Step S211 is marked iris pixel effective in second iris image using binary code label；
Effective iris pixel is labeled as 1, remaining position is labeled as 0；Using two ellipses closest to iris inner and outer boundary, as rainbow
The label of the inner and outer boundary of film；Boundary point is labeled as 1, remaining position is labeled as 0.
The effective iris pixel, without non-irises such as noises and hair, eyelashes, pupil, sclera such as mirror-reflections
Region.
Step S212, the centre mark by boundary oval in the iris marked are the center of pupil, and it includes effective for obtaining
The iris image of iris separator, iris inner and outer boundary label and pupil center's label is as third iris image；The rainbow
Pupil center's point of film image is labeled as 1, remaining position is labeled as 0.
Step S22 instructs the encoding and decoding multitask neural network that the third iris image inputs the full convolution
Practice, obtains the encoding and decoding multitask neural network model of full convolution.
The encoding and decoding multitask neural network of the full convolution includes coding layer and decoding layer.
The coding layer is operated by multiple continuous convolution, ReLu and pondization come coding characteristic, to being originally inputted figure
The mapping graph resolution ratio of picture is continuously reduced, and low resolution mapping graph is obtained.
In the embodiment of the present invention, in order to extract the feature of most judgment in coding layer, by the ASPP of DeepLab or
The pyramid pond module of PSPNet is combined with attention mechanism.
The decoding layer carries out jump connection by layer corresponding with decoding layer, is up-sampled using bilinearity to the coding
Low resolution mapping graph after the continuous diminution of layer carries out resolution ratio recovery, exports four mappings identical with size is originally inputted
Figure, corresponding pupil center's mapping graph, iris inner boundary mapping graph, exterior iris boundary mapping graph and iris segmentation mask mapping graph.
As shown in Figure 4 and Figure 5, the encoding and decoding multitask network of the full convolution of the embodiment of the present invention is with existing VGG-16 model
Based on, full articulamentum is removed, taking remaining layer is the coding layer of multitask network, integrates one in the final stage of coding layer
A attention power module.The attention power module extracts multiple dimensioned contextual information by a context module, utilizes one
Convolutional layer and sigmoid operate to obtain the attention mapping graph of a 3D, and range is between [0,1], finally by this attention
The iris image mapping graph of mapping graph and coding layer final stage carries out being multiplied pixel-by-pixel, then is attached therewith, obtains most
The feature of refinement afterwards.
The context module is the ASPP module of DeepLab or the pyramid pond module of PSPNet.
By aforesaid operations, coding layer is extracted the feature of most essential power of most having any different.Then it is operated using up-sampling,
And jump connection is carried out with the Feature Mapping of coding layer, constantly amplification characteristic figure, up to last acquisition and is originally inputted size
Identical four output mapping graph.As shown in figure 3, the specific implementation of the encoding and decoding multitask network for the full convolution.
Step S221 inputs the third iris image in the encoding and decoding multitask neural network of the full convolution, leads to
Cross propagated forward obtain second pupil center's mapping graph, the second iris inner boundary mapping graph, the second exterior iris boundary mapping graph and
Second iris segmentation mask mapping graph.
Step S222 calculates second pupil center mapping graph, the second iris inner boundary mapping graph, the second iris outside
The error of boundary's mapping graph and the second iris segmentation mask mapping graph and the third iris image, based on the resulting total damage of calculating
Functional value is lost, the encoding and decoding multitask neural network of full convolution is joined using back-propagation algorithm and stochastic gradient descent method
Number updates.
Step S223 repeats step S222, until total loss function calculated value reaches preset condition, is trained
Full convolution encoding and decoding multitask neural network model.
Total loss function includes the friendship of the focal loss for pupil center, the balance for iris inner and outer boundary
Pitch entropy loss and the intersection entropy loss for segmentation.
The focal loss L for pupil centerpupilAs shown in formula (1):
Definition is as shown in formula (2):
Wherein, alpha, gamma is hyper parameter；P={ pj, j=1 ..., | X | }, pjIt is for j-th of pixel of multitask neural network forecast
The probability of pupil center, | X | it is the number of pixels of iris image； For jth in iris image
The label of a pixel, 1 is expressed as true iris pupil center, and 0 is not to be.
The intersection entropy loss L of the balance for iris inner and outer boundaryedgeAs shown in formula (3):
Wherein, Indicate that j-th of pixel belongs to iris inner boundary in iris image
Or the probability of exterior iris boundary, k=1 represent iris inner boundary, k=2 represents exterior iris boundary, and 1 is expressed as true iris
Boundary, 0 is not to be；β is ratio shared by non-edge pixels in iris image, and non-edge pixels includes iris inner boundary and outside
Boundary's pixel.
The intersection entropy loss for segmentation is LsegAs shown in formula (4):
Wherein, S={ sj, j=1 ... | X |, sjIndicate that j-th of pixel belongs to true iris texture in iris image
Probability； J-th of pixel belongs to the label of true iris texture in expression iris image.
As shown in fig. 6, inputting the encoding and decoding multitask neural network of full convolution for single width of embodiment of the present invention iris image
The result of model.
The iris image localization method based on deep learning of second embodiment of the invention, comprising:
Using the step S10- step S20 of the above-mentioned iris segmentation method based on deep learning, the first pupil is obtained
Hole centralizing mapping figure, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and the mapping of the first iris segmentation mask
Figure, and execute following steps:
Step B10, according to the geometry of first pupil center mapping graph and the first iris segmentation mask mapping graph
Relationship obtains the predicted position of pupil center；
Step B20, based on the geometrical relationship between preset pupil, iris, sclera, to the first iris inner and outer boundary
Mapping graph denoising, obtains circle center origin coordinates inside and outside the first iris inside and outside circle radius parameter collection and the first iris；
Step B30 is joined based on radius of circle inside and outside the first iris inner and outer boundary mapping graph and first iris after denoising
Circle center origin coordinates inside and outside manifold and the first iris obtains the match point of iris inside and outside contour using Viterbi algorithm, uses
The circle fitting algorithm of least square is fitted inside and outside circle, obtains Circle Parameters inside and outside iris, completes Iris Location.
In order to be more clearly illustrated to the iris image localization method the present invention is based on deep learning, below with reference to figure
Each step expansion is described in detail in 1 pair of embodiment of the present invention method.
The iris image localization method based on deep learning of an embodiment of the present invention, including step B10- step B30,
Each step is described in detail as follows:
Step B10, according to the geometry of first pupil center mapping graph and the first iris segmentation mask mapping graph
Relationship obtains the predicted position of pupil center.
Threshold segmentation and connected region are carried out to first pupil center mapping graph and the first iris segmentation mask mapping graph
Communicated subarea of the communicated subarea of domain analysis, search and maximum area segmentation mask apart from nearest pupil center, takes institute
Stating communicated subarea central point is the pupil center finally predicted.As shown in fig. 7, being the pupil center location of the embodiment of the present invention
Predictive display figure.
Step B20, based on the geometrical relationship between preset pupil, iris, sclera, to the first iris inner and outer boundary
Mapping graph denoising, obtains circle center origin coordinates inside and outside the first iris inside and outside circle radius parameter collection and the first iris.Such as Fig. 8 institute
Show, is that the iris inner and outer boundary of the embodiment of the present invention denoises, and obtains the flow diagram of the approximate range of iris inside and outside circle.
Step B21, it is maximum to the dicing masks by the center of circle, the pupil center of the predicted position of the pupil center
The maximum distance of the communicated subarea of area is radius, generates limited area.
Step B22 deletes the connected region of the inside and outside exterior iris boundary mapping graph of the limited area, obtains rainbow
Film outer boundary；The connected region for deleting the iris inner boundary mapping graph outside the limited area, obtains iris inner boundary.
Step B23, the minimax distance of the iris inner and outer boundary of pupil center to the refinement is as in the first iris
Exradius parameter set, center coordinate of eye pupil are circle center origin coordinates inside and outside the first iris.
Step B30 is joined based on radius of circle inside and outside the first iris inner and outer boundary mapping graph and first iris after denoising
Circle center origin coordinates inside and outside manifold and the first iris obtains the match point of iris inside and outside contour using Viterbi algorithm, uses
The circle fitting algorithm of least square is fitted inside and outside circle, obtains Circle Parameters inside and outside iris, completes Iris Location.
As shown in figure 9, the result of iris segmentation and positioning is shown for after the completion of all step operations of the embodiment of the present invention
It is intended to.
The iris image acquired under non-controllable scene causes picture quality to be moved back due to illumination, target movement, distance change etc.
Change, the various noises such as mirror-reflection, movement/defocusing blurring, frame be fuzzy is contained, as shown in Fig. 2, the iris image meeting degenerated
So that the segmentation and positioning of iris become difficult, and then influence the precision of iris recognition.
The iris segmentation and positioning system based on deep learning of third embodiment of the invention, including input module,
Multitask neural network module, Threshold segmentation module, pupil position prediction module, denoising module, fitting module, output module；
The input module is configured to obtain iris image to be tested, as the first iris image；
The multitask neural network module is configured to obtain the first pupil center to first iris image processing
Mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and the first iris segmentation mask mapping graph；
The Threshold segmentation module is configured to handle the first iris segmentation mask mapping graph using Threshold segmentation, obtain
Two-value iris segmentation mask artwork is obtained, iris segmentation is completed；
The pupil position prediction module is configured to be covered according to first pupil center mapping graph and the first iris segmentation
The geometrical relationship of Module map figure obtains the predicted position of pupil center；
The denoising module is configured to preset pupil, iris, the geometrical relationship between sclera to first rainbow
The denoising of film inner and outer boundary mapping graph obtains circle center starting inside and outside the first iris inside and outside circle radius parameter collection and the first iris and sits
Mark；
The fitting module is configured in the first iris inner and outer boundary mapping graph and first iris after denoising
Circle center origin coordinates inside and outside exradius parameter set and the first iris obtains iris inside and outside contour using Viterbi algorithm
Match point is fitted inside and outside circle using the circle fitting algorithm of least square, obtains Circle Parameters inside and outside iris, complete Iris Location；
The output module is configured to export Circle Parameters inside and outside the two-value iris segmentation mask artwork and iris.
Person of ordinary skill in the field can be understood that, for convenience and simplicity of description, foregoing description
The specific work process of system and related explanation, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
It should be noted that the iris segmentation and positioning system provided by the above embodiment based on deep learning, only
The example of the division of the above functional modules, in practical applications, it can according to need and by above-mentioned function distribution
Completed by different functional modules, i.e., by the embodiment of the present invention module or step decompose or combine again, for example, on
The module for stating embodiment can be merged into a module, multiple submodule can also be further split into, to complete above description
All or part of function.For module involved in the embodiment of the present invention, the title of step, it is only for distinguish each
Module or step, are not intended as inappropriate limitation of the present invention.
A kind of storage device of fourth embodiment of the invention, wherein being stored with a plurality of program, described program is suitable for by handling
Device is loaded and is executed to realize above-mentioned iris segmentation and localization method based on deep learning.
A kind of processing unit of fifth embodiment of the invention, including processor, storage device；Processor is adapted for carrying out each
Program；Storage device is suitable for storing a plurality of program；Described program is suitable for being loaded by processor and being executed to realize above-mentioned base
In the iris segmentation and localization method of deep learning.
Person of ordinary skill in the field can be understood that, for convenience and simplicity of description, foregoing description
The specific work process and related explanation of storage device, processing unit, can refer to corresponding processes in the foregoing method embodiment,
Details are not described herein.
Those skilled in the art should be able to recognize that, mould described in conjunction with the examples disclosed in the embodiments of the present disclosure
Block, method and step, can be realized with electronic hardware, computer software, or a combination of the two, software module, method and step pair
The program answered can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electric erasable and can compile
Any other form of storage well known in journey ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field is situated between
In matter.In order to clearly demonstrate the interchangeability of electronic hardware and software, in the above description according to function generally
Describe each exemplary composition and step.These functions are executed actually with electronic hardware or software mode, depend on technology
The specific application and design constraint of scheme.Those skilled in the art can carry out using distinct methods each specific application
Realize described function, but such implementation should not be considered as beyond the scope of the present invention.
Term " first ", " second " etc. are to be used to distinguish similar objects, rather than be used to describe or indicate specific suitable
Sequence or precedence.
Term " includes " or any other like term are intended to cover non-exclusive inclusion, so that including a system
Process, method, article or equipment/device of column element not only includes those elements, but also including being not explicitly listed
Other elements, or further include the intrinsic element of these process, method, article or equipment/devices.
The present invention is based on the iris segmentation of deep learning and localization method, system, devices, can apply remote
On the subway import-export system of iris recognition.
Present invention could apply to distant range iris to identify scene.User is allowed to come into the acquisition of subway import-export system first
Region acquires iris image.Farther out due to distance, and user is in movement, therefore collected iris image is faced with fortune
The various noises such as dynamic model pastes, blocks, strabismus.Subway import-export system can use the rainbow provided by the invention based on deep learning
Film segmentation and localization method, system, device, effective effective iris texture of Ground Split, and position the inside and outside circle of iris, Jin Erjin
Row iris normalization, feature extraction, and the iris image registered in result and system is compared, realize user quickly into
Outlet.The case where system can greatly improve the percent of pass of subway user, save personnel's spending and personnel's erroneous judgement.
The present invention is based on the iris segmentation of deep learning and localization method, system, devices, can also be applied to intelligence
The iris solution of mobile phone is locked.
Iris unlock can be widely used in smart phone, facilitated user rapidly to access mobile device, eliminated defeated
Enter the cumbersome process such as password, and whole process be it is safe, there is no the threats for losing password or stolen password.Intelligence
Mobile phone user automatically shoots eye image using built-in or external near-infrared or visible image capturing head, but the figure obtained
As that can have many noise jammings, such as mirror-reflection, motion blur etc., the rainbow provided by the invention based on deep learning is utilized
Film segmentation and localization method, system, device, can effective Ground Split effective iris texture, and position the inside and outside circle of iris, into
And iris normalization, feature extraction are carried out, and the iris image registered in result and cell phone system is compared, realize user
Quickly unlock.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this field
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from this
Under the premise of the principle of invention, those skilled in the art can make equivalent change or replacement to the relevant technologies feature, these
Technical solution after change or replacement will fall within the scope of protection of the present invention.
Claims (11)
1. a kind of iris segmentation method based on deep learning characterized by comprising
Step S10 obtains iris image to be processed, as the first iris image；
Step S20 carries out the trained full convolution encoding and decoding multitask neural network model of first iris image input
Propagated forward obtains first pupil center's mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and
One iris segmentation mask mapping graph；
Step S30 handles the first iris segmentation mask mapping graph using Threshold segmentation, obtains two-value iris segmentation mask
Figure completes iris segmentation；
The trained full convolution encoding and decoding multitask neural network model, acquisition modes are as follows:
Step S21 obtains iris image and second iris image is marked as the second iris image, obtains third
Iris image；Attention mechanism is incorporated, the encoding and decoding multitask neural network of full convolution is constructed；
The encoding and decoding multitask neural network that the third iris image inputs the full convolution is trained, obtains by step S22
Obtain the encoding and decoding multitask neural network model of trained full convolution.
2. the iris segmentation method according to claim 1 based on deep learning, which is characterized in that step S21 is " right
Second iris image is marked ", it the steps include:
Step S211 is marked iris pixel effective in second iris image using binary code label；It uses
Two ellipses closest to iris inner and outer boundary, the label of the inner and outer boundary as iris；
Step S212, the centre mark by boundary oval in the iris marked are the center of pupil, obtain effective iris and distinguish
The iris image of tag image, iris inner and outer boundary tag image and pupil center's tag image is as third iris image.
3. the iris segmentation method according to claim 1 based on deep learning, which is characterized in that the full convolution
Encoding and decoding multitask neural network include coding layer and decoding layer；
The coding layer is operated by multiple continuous convolution, ReLu and pondization come coding characteristic, to original input picture
Mapping graph resolution ratio is continuously reduced, and low resolution mapping graph is obtained；
The decoding layer carries out jump connection by layer corresponding with decoding layer, is connected using bilinearity up-sampling to the coding layer
Low resolution mapping graph after continuous diminution carries out resolution ratio recovery, exports four mapping graphs identical with size is originally inputted, right
Answer pupil center's mapping graph, iris inner boundary mapping graph, exterior iris boundary mapping graph and iris segmentation mask mapping graph.
4. the iris segmentation method according to claim 1 based on deep learning, which is characterized in that in step S22
Described " being trained the encoding and decoding multitask neural network that the third iris image inputs the full convolution ", step
Are as follows:
Step S221 inputs the third iris image in the encoding and decoding multitask neural network of the full convolution, by preceding
Second pupil center's mapping graph, the second iris inner boundary mapping graph, the second exterior iris boundary mapping graph and second are obtained to propagation
Iris segmentation mask mapping graph；
Step S222 calculates second pupil center mapping graph, the second iris inner boundary mapping graph, the second exterior iris boundary and reflects
The error for penetrating figure and the second iris segmentation mask mapping graph and the third iris image, based on the resulting total loss letter of calculating
Numerical value carries out parameter more using back-propagation algorithm and stochastic gradient descent method to the encoding and decoding multitask neural network of full convolution
Newly；
Step S223 repeats step S222, and until total loss function calculated value reaches preset condition, acquisition is trained complete
The encoding and decoding multitask neural network model of convolution.
5. the iris segmentation method according to claim 4 based on deep learning, which is characterized in that total damage
Lose function include for pupil center focal loss, for iris inner and outer boundary balance intersection entropy loss and for segmentation
Intersection entropy loss；
The focal loss L for pupil centerpupilAre as follows:
Wherein, alpha, gamma is hyper parameter；P={ pj, j=1 ..., | X | }, pjIt is pupil for j-th of pixel of multitask neural network forecast
The probability at center, | X | it is the number of pixels of iris image； For j-th of pixel in iris image
Label, 1 is expressed as true iris pupil center, and 0 is is not；
The intersection entropy loss L of the balance for iris inner and outer boundaryedgeAre as follows:
Wherein, Indicate iris image in j-th of pixel belong to iris inner boundary or
The probability of exterior iris boundary, k=1 represent iris inner boundary, and k=2 represents exterior iris boundary, and 1 is expressed as true iris boundary,
0 is not to be；β is ratio shared by non-edge pixels in iris image, and non-edge pixels includes iris inner boundary and outer boundary picture
Element；
The intersection entropy loss for segmentation is LsegAre as follows:
Wherein, S={ sj, j=1 ... | X |, sjIndicate that j-th of pixel belongs to the general of true iris texture in iris image
Rate； J-th of pixel belongs to the label of true iris texture in expression iris image.
6. a kind of iris image localization method based on deep learning characterized by comprising
Using the step S10- step S20 of the iris segmentation method described in claim 1-5 based on deep learning, obtain
First pupil center's mapping graph, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and the first iris segmentation are covered
Module map figure, and execute following steps:
Step B10, according to the geometrical relationship of first pupil center mapping graph and the first iris segmentation mask mapping graph,
Obtain the predicted position of pupil center；
Step B20 maps the first iris inner and outer boundary based on the geometrical relationship between preset pupil, iris, sclera
Figure denoising, obtains circle center origin coordinates inside and outside the first iris inside and outside circle radius parameter collection and the first iris；
Step B30, based on the first iris inner and outer boundary mapping graph and the first iris inside and outside circle radius parameter collection after denoising
And the first circle center origin coordinates inside and outside iris, the match point of iris inside and outside contour is obtained using Viterbi algorithm, uses minimum
The two circle fitting algorithm fitting inside and outside circles multiplied, obtain Circle Parameters inside and outside iris, complete Iris Location.
7. the iris image localization method according to claim 6 based on deep learning, which is characterized in that in step B10
It " according to the geometrical relationship of first pupil center mapping graph and the first iris segmentation mask mapping graph, obtains in pupil
The predicted position of the heart ", method are as follows:
Threshold segmentation and connected region point are carried out to first pupil center mapping graph and the first iris segmentation mask mapping graph
Communicated subarea of the communicated subarea of analysis, search and maximum area segmentation mask apart from nearest pupil center, takes the company
Logical subregion central point is the pupil center finally predicted.
8. the iris image localization method according to claim 6 or 7 based on deep learning, which is characterized in that step B20
In " based on the geometrical relationship between preset pupil, iris, sclera, the first iris inner and outer boundary mapping graph is denoised, is obtained
Circle center origin coordinates inside and outside to the first iris inside and outside circle radius parameter collection and the first iris ", the steps include:
Step B21, using the predicted position of the pupil center as the center of circle, the pupil center to the dicing masks maximum area
Communicated subarea maximum distance be radius, generate limited area；
Step B22 deletes the inside and outside exterior iris boundary mapping graph connected region of the limited area, obtains outside iris
Boundary；The iris inner boundary mapping graph connected region outside the limited area is deleted, iris inner boundary is obtained；
The minimax distance of step B23, the iris inner and outer boundary of pupil center to the refinement are used as the first iris inside and outside circle
Radius parameter collection, center coordinate of eye pupil are circle center origin coordinates inside and outside the first iris.
9. a kind of iris segmentation and positioning system based on deep learning, which is characterized in that including input module, multitask
Neural network module, Threshold segmentation module, pupil position prediction module, denoising module, fitting module, output module；
The input module is configured to obtain iris image to be tested, as the first iris image；
The multitask neural network module is configured to obtain the mapping of the first pupil center to first iris image processing
Figure, the first iris inner boundary mapping graph, the first exterior iris boundary mapping graph and the first iris segmentation mask mapping graph；
The Threshold segmentation module is configured to handle the first iris segmentation mask mapping graph using Threshold segmentation, obtains two
It is worth iris segmentation mask artwork, completes iris segmentation；
The pupil position prediction module is configured to be reflected according to first pupil center mapping graph and the first iris segmentation mask
The geometrical relationship for penetrating figure obtains the predicted position of pupil center；
The denoising module is configured to preset pupil, iris, the geometrical relationship between sclera in first iris
The denoising of outer boundary mapping graph, obtains circle center origin coordinates inside and outside the first iris inside and outside circle radius parameter collection and the first iris；
The fitting module, the first iris inner and outer boundary mapping graph and the first iris inside and outside circle after being configured to denoising
Circle center origin coordinates inside and outside radius parameter collection and the first iris obtains the fitting of iris inside and outside contour using Viterbi algorithm
Point is fitted inside and outside circle using the circle fitting algorithm of least square, obtains Circle Parameters inside and outside iris, complete Iris Location；
The output module is configured to export Circle Parameters inside and outside the two-value iris segmentation mask artwork and iris.
10. a kind of storage device, wherein being stored with a plurality of program, which is characterized in that described program is suitable for by processor load simultaneously
It executes to realize the described in any item iris segmentations and localization method based on deep learning of claim 1-8.
11. a kind of processing unit, including
Processor is adapted for carrying out each program；And
Storage device is suitable for storing a plurality of program；
It is characterized in that, described program is suitable for being loaded by processor and being executed to realize:
The described in any item iris segmentations and localization method based on deep learning of claim 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910002663.XA CN109815850B (en) | 2019-01-02 | 2019-01-02 | Iris image segmentation and positioning method, system and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910002663.XA CN109815850B (en) | 2019-01-02 | 2019-01-02 | Iris image segmentation and positioning method, system and device based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109815850A true CN109815850A (en) | 2019-05-28 |
| CN109815850B CN109815850B (en) | 2020-11-10 |
Family
ID=66603779
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910002663.XA Active CN109815850B (en) | 2019-01-02 | 2019-01-02 | Iris image segmentation and positioning method, system and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109815850B (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458833A (en) * | 2019-08-15 | 2019-11-15 | 腾讯科技（深圳）有限公司 | Medical image processing method, medical supply and storage medium based on artificial intelligence |
| CN110458016A (en) * | 2019-07-08 | 2019-11-15 | 北京虹安翔宇信息科技有限公司 | A kind of processing method of the iris recognition image based on high pass chip |
| CN110490083A (en) * | 2019-07-23 | 2019-11-22 | 苏州国科视清医疗科技有限公司 | A kind of pupil accurate detecting method based on fast human-eye semantic segmentation network |
| CN110738663A (en) * | 2019-09-06 | 2020-01-31 | 上海衡道医学病理诊断中心有限公司 | Double-domain adaptive module pyramid network and unsupervised domain adaptive image segmentation method |
| CN110738107A (en) * | 2019-09-06 | 2020-01-31 | 上海衡道医学病理诊断中心有限公司 | microscopic image recognition and segmentation method with model migration function |
| CN110991415A (en) * | 2019-12-21 | 2020-04-10 | 武汉中海庭数据技术有限公司 | Structural target high-precision segmentation method, electronic equipment and storage medium |
| CN111161276A (en) * | 2019-11-27 | 2020-05-15 | 天津中科智能识别产业技术研究院有限公司 | Iris normalized image forming method |
| CN111588469A (en) * | 2020-05-18 | 2020-08-28 | 四川大学华西医院 | Ophthalmic robot end effector guidance and positioning system |
| CN111738122A (en) * | 2020-06-12 | 2020-10-02 | Oppo广东移动通信有限公司 | Image processing method and related device |
| CN111938567A (en) * | 2020-07-09 | 2020-11-17 | 上海交通大学 | Deep learning-based ophthalmologic parameter measurement method, system and equipment |
| CN112287872A (en) * | 2020-11-12 | 2021-01-29 | 北京建筑大学 | Iris image segmentation, positioning and normalization method based on multitask neural network |
| CN112651328A (en) * | 2020-12-23 | 2021-04-13 | 浙江中正智能科技有限公司 | Iris segmentation method based on geometric position relation loss function |
| CN112668472A (en) * | 2020-12-28 | 2021-04-16 | 中国科学院自动化研究所 | Iris image feature extraction method, system and device based on federal learning |
| CN113253850A (en) * | 2021-07-05 | 2021-08-13 | 中国科学院西安光学精密机械研究所 | Multitask cooperative operation method based on eye movement tracking and electroencephalogram signals |
| CN113537111A (en) * | 2021-07-26 | 2021-10-22 | 南京信息工程大学 | Iris segmentation method based on double-branch deep convolutional network |
| CN113536968A (en) * | 2021-06-25 | 2021-10-22 | 天津中科智能识别产业技术研究院有限公司 | Method for automatically acquiring boundary coordinates of inner circle and outer circle of iris |
| CN113689385A (en) * | 2021-07-29 | 2021-11-23 | 天津中科智能识别产业技术研究院有限公司 | Method, device and equipment for automatically dividing inner and outer circle boundaries of iris and storage medium |
| CN113706469A (en) * | 2021-07-29 | 2021-11-26 | 天津中科智能识别产业技术研究院有限公司 | Iris automatic segmentation method and system based on multi-model voting mechanism |
| CN113706470A (en) * | 2021-07-29 | 2021-11-26 | 天津中科智能识别产业技术研究院有限公司 | Iris image segmentation method and device, electronic equipment and storage medium |
| CN114638879A (en) * | 2022-03-21 | 2022-06-17 | 四川大学华西医院 | Medical pupil size measuring system |
| CN115797632A (en) * | 2022-12-01 | 2023-03-14 | 北京科技大学 | Image segmentation method based on multi-task learning |
| CN116110113A (en) * | 2022-11-15 | 2023-05-12 | 南昌航空大学 | Iris recognition method based on deep learning |
| CN117316437A (en) * | 2023-11-29 | 2023-12-29 | 首都医科大学附属北京安定医院 | Pain level prediction method, system and equipment based on pupil change |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN106778664A (en) * | 2016-12-29 | 2017-05-31 | 天津中科智能识别产业技术研究院有限公司 | The dividing method and its device of iris region in a kind of iris image |
-
2019
- 2019-01-02 CN CN201910002663.XA patent/CN109815850B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN106778664A (en) * | 2016-12-29 | 2017-05-31 | 天津中科智能识别产业技术研究院有限公司 | The dividing method and its device of iris region in a kind of iris image |
Non-Patent Citations (5)
| Title |
|---|
| CUNJIAN CHEN等: "A Multi-Task Convolutional Neural Network for Joint Iris Detection and Presentation Attack Detection", 《IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION》 * |
| NIANFENG LIU等: "Accurate Iris Segmentation in Non-cooperative Environments Using Fully Convolutional Networks", 《IEEE》 * |
| TSUNG-YI LIN等: "Focal Loss for Dense Object Detection", 《ARXIV》 * |
| ZIJING ZHAO等: "An Accurate Iris Segmentation Framework under Relaxed Imaging Constraints using Total Variation Model", 《2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
| 姜华 等: "基于CS-LBP与自适应神经网络的虹膜识别", 《东北师大学报（自然科学版）》 * |
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458016A (en) * | 2019-07-08 | 2019-11-15 | 北京虹安翔宇信息科技有限公司 | A kind of processing method of the iris recognition image based on high pass chip |
| CN110490083A (en) * | 2019-07-23 | 2019-11-22 | 苏州国科视清医疗科技有限公司 | A kind of pupil accurate detecting method based on fast human-eye semantic segmentation network |
| CN110458833A (en) * | 2019-08-15 | 2019-11-15 | 腾讯科技（深圳）有限公司 | Medical image processing method, medical supply and storage medium based on artificial intelligence |
| CN110458833B (en) * | 2019-08-15 | 2023-07-11 | 腾讯科技（深圳）有限公司 | Medical image processing method, medical device and storage medium based on artificial intelligence |
| CN110738663A (en) * | 2019-09-06 | 2020-01-31 | 上海衡道医学病理诊断中心有限公司 | Double-domain adaptive module pyramid network and unsupervised domain adaptive image segmentation method |
| CN110738107A (en) * | 2019-09-06 | 2020-01-31 | 上海衡道医学病理诊断中心有限公司 | microscopic image recognition and segmentation method with model migration function |
| CN111161276A (en) * | 2019-11-27 | 2020-05-15 | 天津中科智能识别产业技术研究院有限公司 | Iris normalized image forming method |
| CN111161276B (en) * | 2019-11-27 | 2023-04-18 | 天津中科智能识别产业技术研究院有限公司 | Iris normalized image forming method |
| CN110991415A (en) * | 2019-12-21 | 2020-04-10 | 武汉中海庭数据技术有限公司 | Structural target high-precision segmentation method, electronic equipment and storage medium |
| CN111588469A (en) * | 2020-05-18 | 2020-08-28 | 四川大学华西医院 | Ophthalmic robot end effector guidance and positioning system |
| CN111588469B (en) * | 2020-05-18 | 2021-02-02 | 四川大学华西医院 | Ophthalmic robot end effector guidance and positioning system |
| CN111738122A (en) * | 2020-06-12 | 2020-10-02 | Oppo广东移动通信有限公司 | Image processing method and related device |
| CN111738122B (en) * | 2020-06-12 | 2023-08-22 | Oppo广东移动通信有限公司 | Image processing method and related device |
| CN111938567A (en) * | 2020-07-09 | 2020-11-17 | 上海交通大学 | Deep learning-based ophthalmologic parameter measurement method, system and equipment |
| CN111938567B (en) * | 2020-07-09 | 2021-10-22 | 上海交通大学 | Deep learning-based ophthalmologic parameter measurement method, system and equipment |
| CN112287872B (en) * | 2020-11-12 | 2022-03-25 | 北京建筑大学 | Iris image segmentation, positioning and normalization method based on multitask neural network |
| CN112287872A (en) * | 2020-11-12 | 2021-01-29 | 北京建筑大学 | Iris image segmentation, positioning and normalization method based on multitask neural network |
| CN112651328A (en) * | 2020-12-23 | 2021-04-13 | 浙江中正智能科技有限公司 | Iris segmentation method based on geometric position relation loss function |
| WO2022142060A1 (en) * | 2020-12-28 | 2022-07-07 | 中国科学院自动化研究所 | Iris image feature extraction method and system based on federated learning, and apparatus |
| CN112668472B (en) * | 2020-12-28 | 2021-08-31 | 中国科学院自动化研究所 | Iris image feature extraction method, system and device based on federal learning |
| CN112668472A (en) * | 2020-12-28 | 2021-04-16 | 中国科学院自动化研究所 | Iris image feature extraction method, system and device based on federal learning |
| CN113536968B (en) * | 2021-06-25 | 2022-08-16 | 天津中科智能识别产业技术研究院有限公司 | Method for automatically acquiring boundary coordinates of inner and outer circles of iris |
| CN113536968A (en) * | 2021-06-25 | 2021-10-22 | 天津中科智能识别产业技术研究院有限公司 | Method for automatically acquiring boundary coordinates of inner circle and outer circle of iris |
| CN113253850A (en) * | 2021-07-05 | 2021-08-13 | 中国科学院西安光学精密机械研究所 | Multitask cooperative operation method based on eye movement tracking and electroencephalogram signals |
| CN113537111A (en) * | 2021-07-26 | 2021-10-22 | 南京信息工程大学 | Iris segmentation method based on double-branch deep convolutional network |
| CN113689385B (en) * | 2021-07-29 | 2023-10-20 | 天津中科智能识别产业技术研究院有限公司 | Automatic iris inner and outer circle boundary segmentation method, device, equipment and storage medium |
| CN113706470A (en) * | 2021-07-29 | 2021-11-26 | 天津中科智能识别产业技术研究院有限公司 | Iris image segmentation method and device, electronic equipment and storage medium |
| CN113706469A (en) * | 2021-07-29 | 2021-11-26 | 天津中科智能识别产业技术研究院有限公司 | Iris automatic segmentation method and system based on multi-model voting mechanism |
| CN113689385A (en) * | 2021-07-29 | 2021-11-23 | 天津中科智能识别产业技术研究院有限公司 | Method, device and equipment for automatically dividing inner and outer circle boundaries of iris and storage medium |
| CN113706470B (en) * | 2021-07-29 | 2023-12-15 | 天津中科智能识别产业技术研究院有限公司 | Iris image segmentation method and device, electronic equipment and storage medium |
| CN113706469B (en) * | 2021-07-29 | 2024-04-05 | 天津中科智能识别产业技术研究院有限公司 | Iris automatic segmentation method and system based on multi-model voting mechanism |
| CN114638879A (en) * | 2022-03-21 | 2022-06-17 | 四川大学华西医院 | Medical pupil size measuring system |
| CN116110113A (en) * | 2022-11-15 | 2023-05-12 | 南昌航空大学 | Iris recognition method based on deep learning |
| CN115797632A (en) * | 2022-12-01 | 2023-03-14 | 北京科技大学 | Image segmentation method based on multi-task learning |
| CN115797632B (en) * | 2022-12-01 | 2024-02-09 | 北京科技大学 | Image segmentation method based on multi-task learning |
| CN117316437A (en) * | 2023-11-29 | 2023-12-29 | 首都医科大学附属北京安定医院 | Pain level prediction method, system and equipment based on pupil change |
| CN117316437B (en) * | 2023-11-29 | 2024-03-08 | 首都医科大学附属北京安定医院 | Pain level prediction method, system and equipment based on pupil change |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109815850B (en) | 2020-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109815850A (en) | Iris segmentation and localization method, system, device based on deep learning | |
| CN106778664B (en) | Iris image iris area segmentation method and device | |
| CN111274916B (en) | Face recognition method and face recognition device | |
| Nakamura et al. | Scene text eraser | |
| CN101359365B (en) | Iris positioning method based on maximum between-class variance and gray scale information | |
| CN101261677B (en) | New method-feature extraction layer amalgamation for face | |
| Shams et al. | Iris recognition based on LBP and combined LVQ classifier | |
| Morris | A pyramid CNN for dense-leaves segmentation | |
| CN101201893A (en) | Iris recognizing preprocessing method based on grey level information | |
| CN102844766A (en) | Human eyes images based multi-feature fusion identification method | |
| CN112837344B (en) | Target tracking method for generating twin network based on condition countermeasure | |
| CN107169479A (en) | Intelligent mobile equipment sensitive data means of defence based on fingerprint authentication | |
| CN114241548A (en) | Small target detection algorithm based on improved YOLOv5 | |
| CN113361495A (en) | Face image similarity calculation method, device, equipment and storage medium | |
| CN106611158A (en) | Method and equipment for obtaining human body 3D characteristic information | |
| CN111539320B (en) | Multi-view gait recognition method and system based on mutual learning network strategy | |
| CN114821682A (en) | Multi-sample mixed palm vein identification method based on deep learning algorithm | |
| Ngxande et al. | Detecting inter-sectional accuracy differences in driver drowsiness detection algorithms | |
| Hu et al. | Computer vision based method for severity estimation of tea leaf blight in natural scene images | |
| CN111666813B (en) | Subcutaneous sweat gland extraction method of three-dimensional convolutional neural network based on non-local information | |
| CN116631068A (en) | Palm vein living body detection method based on deep learning feature fusion | |
| CN116524549A (en) | Method for positioning key points and ROI (region of interest) of back or palm vein image based on improved UNet | |
| Sinha et al. | Iris segmentation using deep neural networks | |
| Zabihi et al. | Vessel extraction of conjunctival images using LBPs and ANFIS | |
| Hashim et al. | Fast Iris localization based on image algebra and morphological operations |
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 |