CN1760887A - The robust features of iris image extracts and recognition methods - Google Patents
The robust features of iris image extracts and recognition methods Download PDFInfo
- Publication number
- CN1760887A CN1760887A CN 200410081184 CN200410081184A CN1760887A CN 1760887 A CN1760887 A CN 1760887A CN 200410081184 CN200410081184 CN 200410081184 CN 200410081184 A CN200410081184 A CN 200410081184A CN 1760887 A CN1760887 A CN 1760887A
- Authority
- CN
- China
- Prior art keywords
- iris
- wave filter
- image
- iris image
- multipole
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 51
- 239000000284 extract Substances 0.000 title claims abstract description 13
- 230000005405 multipole Effects 0.000 claims abstract description 52
- 238000001914 filtration Methods 0.000 claims abstract description 35
- 239000004744 fabric Substances 0.000 claims abstract description 26
- 238000010606 normalization Methods 0.000 claims abstract description 25
- 210000001747 pupil Anatomy 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 10
- 239000000203 mixture Substances 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 230000013011 mating Effects 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 241000790351 Iris latifolia Species 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000005484 gravity Effects 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 230000002093 peripheral effect Effects 0.000 claims description 2
- 238000011002 quantification Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 6
- 238000005286 illumination Methods 0.000 abstract description 4
- 238000012545 processing Methods 0.000 abstract description 2
- 210000000744 eyelid Anatomy 0.000 description 9
- 210000000720 eyelash Anatomy 0.000 description 8
- 230000008878 coupling Effects 0.000 description 5
- 238000010168 coupling process Methods 0.000 description 5
- 238000005859 coupling reaction Methods 0.000 description 5
- 230000000903 blocking effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 239000003550 marker Substances 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000012850 discrimination method Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 108010028295 histidylhistidine Proteins 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 210000003786 sclera Anatomy 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Abstract
The present invention relates to the digital image processing techniques field, a kind of authentication identifying method that extracts based on the iris image robust features comprises step: the iris image pre-service; Use the multipole wave filter normalization iris image is carried out airspace filter; The result of filtering is carried out binary-coding, make up the proper vector of this image; Calculate the Weighted H amming distance between the proper vector of two width of cloth iris images.The present invention adopts novel " multipole wave filter " to be used for extracting the classified information that iris texture contains; The filtering result is carried out the robust coding; The similarity of measuring two width of cloth images according to Weighted H amming distance.The present invention has the little advantage of strong robustness, memory space that computing velocity is fast, accuracy of identification is high, resist noise and illumination variation.The present invention needing can be used for being particularly useful for calculating storage resources platforms more in short supply such as PDA, mobile phone and embedded system and carrying out authentication in many application systems of identification.
Description
Technical field
The robust features that the present invention relates to Digital Image Processing, pattern-recognition, computer vision and information coding technique field, particularly a kind of iris image extracts and recognition methods.
Background technology
In human society because the different division of labor, everyone is playing the part of different roles, bear different obligatioies simultaneously, enjoy different rights, many important social activitieies all need to confirm participant's identity, so identification is a unfailing topic.Traditional identification mode mainly comprises based on knowledge (as password, password) or based on the combine method of (as bank card) of marker (as key, IC-card) or both.The shortcoming of these methods is that password may be cracked, may pass out of mind, and marker may be lost, may be forged.In the higher occasion of security requirement, these and the identity identifying method that does not rely on people's unique characteristics can't satisfy the demands.Along with the development of computer and network technologies, shown wide application prospect based on the personal identification method of biological characteristic (as fingerprint, iris, face picture).These physiological characteristics or behavioural characteristic have " that the people respectively has is different, throughout one's life constant, carry " etc. advantage, for people's exploitation and use the smart identity identification system of high reliability to provide may.Iris is as a visible internal in outside in various biological characteristics, and its image has abundant texture and minutia can be used for identification.Iris recognition has characteristics such as uniqueness height, stability are strong, noncontact.Iris recognition technology has been successfully applied to the identity of occasions such as airport, customs, bank, school, hospital and has differentiated.
Mobile electronic device (as mobile phone, the PDA of band digital camera) has begun to enter into huge numbers of families in recent years, add the development of wireless telecommunications, people have begun to receive Email, carry out ecommerce, participate in stock market's dealing, store private secret with these mobile devices.These occasions all need to carry out authentication, and adopting iris recognition will be a safe and reliable selection.Because be subjected to the restriction of volume and energy, the computational resource of these equipment, storage resources all are limited, and the quality of the iris image that collects can not be guaranteed.Deficiency that so speed is fast, precision is high, strong robustness, Algorithm of Iris Recognition that memory space is little can remedy hardware device preferably.
Summary of the invention
The robust features that the objective of the invention is to propose a kind of iris image extracts and recognition methods, utilizes the relative half-tone information of zones of different in the iris image to determine the method for people's identity.
For achieving the above object, the robust features of iris image extracts with recognition methods and comprises step:
The iris image pre-service;
Use the multipole wave filter normalization iris image under the polar coordinates is carried out airspace filter;
The result of filtering is carried out binary-coding, make up the proper vector of iris image;
Calculate the Weighted H amming distance (the weighting template is offline created in advance by the statistical learning of large sample) between the proper vector of two width of cloth iris images;
According to the Weighted H amming that calculates gained apart from judging that two above-mentioned width of cloth iris images are whether from same people's same eyes.
Iris discrimination method of the present invention has designed novel " multipole wave filter " but has extracted distinguishing characteristic in the iris texture, the advantage of this wave filter is the configuration simple and flexible, can portray contrast directional information stable in the iris image, be subjected to interference of noise little, different illumination, contrast and focus level etc. are to the influence of feature coding in the time of overcoming imaging; Adopt the binary-coding of robust that characteristic information is carried out qualitative description, though the lost part information content, what remain all is robustness and the extremely strong essential information of stability, has also saved storage space, has been convenient to coupling; But the feature of a width of cloth iris image has up to a hundred variable freedoms, enough carries out the identification of large-scale crowd; Adopt Weighted H amming distance can when making full use of each area information of iris, reduce the influence that eyelashes, eyelid etc. block as similarity measurement.The present invention can be used for needing authentication, carries out in many application systems of security strick precaution, as ecommerce, E-Government, electronics military task, electronic police affair etc.Algorithm of the present invention is low for the performance requirement of hardware, and calculating is very simple, is easy to be written as software or realizes (as programmable hardware unit such as FPGA) with hardware.So being particularly useful for calculating storage resources platforms more in short supply such as PDA, mobile phone and embedded system, the present invention carries out authentication.
Iris authentication system is generally operational under the two states: registration mode and recognition mode.In registration mode, validated user is submitted the iris feature template of oneself to system; At recognition mode, system determines by the template of contrast storage and the feature of the interim iris image of gathering of user whether this user identity is legal, no matter be registration mode or recognition mode, iris identification method all will carry out image pre-service and feature extraction, also will carry out characteristic matching under recognition mode.The method that the present invention proposes will be mainly used in iris feature and extract and coupling.
Described " multipole wave filter " constituted jointly by the low-pass filter (for example Gauss's template) that differ, positive and negatively differ in a plurality of positions, yardstick differs and shape differs, each low-pass filter is called one " utmost point ", the coefficient summation of all wave filters is 0, promptly " multipole wave filter " is a kind of form differential filter flexibly, and the first order derivative wave filter of Gabor wave filter, Gaussian function, the second derivative wave filter of Gaussian function, general differential filters such as wavelet filter also belong to the special circumstances of " multipole wave filter ".
The low-pass filter of described composition " multipole wave filter " can be square or box-shaped wave filter (box filter), promptly to the average wave filter of filtering image zone summation.
To the airspace filter process of normalization iris image be exactly: the multipole wave filter is roamed on the normalization iris image, and the gray-scale value of the image-region that is covered by the multipole wave filter and the window coefficient of wave filter carry out behind the dot product filtering result as the image pixel of corresponding multipole template center.
Sign symbol according to above-mentioned filtering result is carried out binary-coding to filtered image-region, for example: if the filtering result is greater than 0 then the feature coding of regional center pixel is 1; If the filtering result is smaller or equal to 0 then the feature coding of regional center pixel is 0, all at last feature codings are composed in series the proper vector of this iris image, perhaps be expressed as binary matrix form, the result of " multipole wave filter " filtering also can be by other quantification gradation coding.
Proper vector or the eigenmatrix of describing a width of cloth iris image can be proper vector or the eigenmatrixes that is constituted jointly by a plurality of different multipole wave filter coding results.
Can be down-sampled to proper vector to reduce the length of proper vector according to actual conditions, save storage space and match time.
Judge that whether two width of cloth iris images depend on from same people's same eyes whether their the Weighted H amming distance of proper vector is enough little.
The weighting template is offline created in advance by the statistical learning of large sample, and the weights of each condition code are decided according to its discrimination: discrimination is high more, and corresponding weights are also big more.
Description of drawings
Fig. 1 is that the robust features of iris image extracts and the recognition methods FB(flow block).
Fig. 2 is an iris image pre-service synoptic diagram.
Fig. 3 is two examples of multipole wave filter.
The weighting template of Fig. 4 for obtaining by off-line learning.
The distribution of the matching result that Fig. 5 tests on fairly large iris image database for the present invention.
Embodiment
The robust features that the present invention proposes a kind of iris image of novelty extracts and recognition methods its FB(flow block) such as Fig. 1.
With comparing of current other iris identification methods, novelty of the present invention mainly is: 1) utilize the multipole wave filter to extract the information of three different levels in the iris image: each iris feature representation be the zone at center and the relevant nature of multipole wave filter with certain image pixel, 1 represents positive correlation, and 0 represents negative correlation; The smothing filtering result of corresponding small neighbourhood has represented the average gray information that this is regional in each utmost point (low-pass filter) in the multipole wave filter and the image-region, and its result has partly determined the value of condition code; All at last condition codes are formed the proper vector of describing the iris image Global Information.2) each in the multipole wave filter extremely can flexible configuration, can select diverse location, different scale and difformity, and the multipole wave filter can deeply be portrayed various complexity in the iris image and trickle structure (referring to Fig. 3) like this.3) iris feature is expressed as the relatively half-tone information rather than the quantitative or absolute half-tone information qualitatively of a plurality of image-regions, such benefit is the linear-scale variation that characteristics of image is independent of image gray levels, can resist of the influence of different imaging circumstances such as illumination, contrast, sharpness to recognition performance, reduction has improved robustness to the requirement of picture quality.4) can arrange the relative position of contrast images piece flexibly, and not require adjacent.Compare with respect to neighborhood, internal information (because the similarity height of adjacent image piece of iris image can be more effectively excavated in non-neighborhood comparison, and has bigger independence between the non-conterminous image block, have more entropy, provide more information), noise (difference of non-neighbour's comparison is generally bigger, and bigger noise tolerance is arranged) is resisted on robust ground more.5) binary-coding is economized storage space, is convenient to coupling.6) adopt Weighted H amming distance rather than simple Hamming distance from the similarity of measuring between iris image, can not require the influence that reduces them under the situation that detection eyelashes and eyelid etc. block as much as possible.
The robust features of the iris image that the present invention proposes extracts with recognition methods and comprises seven key step (see figure 1)s:
1. iris image acquiring S1 is under the abundant irradiation of near infrared light of 800-900nm at wavelength, just can photograph the iris image clearly that is rich in detail textures information with common CCD or CMOS camera;
2. iris image pre-service S2.Not only comprise iris in the iris image, also have pupil, sclera, eyelid and eyelashes etc.The all very approaching circle of the outline of pupil and iris, so the pretreated task of iris image is to find the coordinate and the radius of the circle of match pupil and iris boundary, then the iris annulus normalized to the rectangular area of fixed measure, i.e. Iris Location and normalization;
I. Iris Location
The gray scale of human eye pupil is far below the peripheral region, isolate pupil region so can use threshold method, center of gravity that then should the zone is gone to the edge of match pupil as preliminary pupil center with variable-sized template near this point, best fitting result is exactly the result of pupil location.The center of iris is near the center of pupil, so can make center and the radius that uses the same method and find iris.Fig. 2 (b) is to the example after the Iris Location among Fig. 2 (a), and wherein Bai Se circle is represented the outer boundary of pupil and iris after the match.
II. normalization
In the mode of bilinear interpolation, the iris annulus of having good positioning can be carried out the rectangular area of spatial alternation to a fixed measure.Fig. 2 (c) is the result after the iris normalization.Horizontal direction normalizes to 0 °～360 ° corresponding to the angle direction (being positive dirction counterclockwise) of original iris image, and vertical direction normalizes to 0～1 corresponding to the radial direction of original iris image;
3. adopt " multipole wave filter " to carry out feature extraction S3, use the multipole wave filter normalization iris image of pre-service output is carried out airspace filter: the multipole wave filter is roamed on the normalization iris image, and the gray-scale value of the image-region that is covered by the multipole wave filter and the window coefficient of wave filter carry out behind the dot product filtering result as the image pixel of corresponding multipole template center;
4. robust coding S4 carries out binary-coding according to the filtering result's of previous step sign symbol to filtered image-region, for example: if the filtering result is greater than 0 then the feature coding of regional center pixel is 1; If the filtering result is smaller or equal to 0 then the feature coding of regional center pixel is 0.The composition characteristic vector of at last all condition codes being joined together.Can be down-sampled to proper vector to reduce the length of proper vector according to actual conditions, save storage space and match time.If registration process is kept at proper vector in the template database; If identifying is then with the input of proper vector as matching engine.
5. characteristic matching S5, when biological identification system operates under the certification mode, the user will state the identity of oneself, and system finds out this people according to his statement from database template characteristic vector sum input feature value mates, and promptly proves his his said that people really.When biological identification system operates in recognition mode following time, the user need not tell system's " Who Am I ", system according to he input feature vector and database in all feature templates mate one by one.All identity differential modes are all to mate in twos.Because the top of the iris region in iris image is subjected to blocking of eyelashes and eyelid easily, the bottom is subjected to blocking of eyelid easily, resulting condition code is not very reliable when multipole wave filter and these regional filtering, obtains rational recognition result so the condition code of diverse location must be provided with different weights as far as possible.Based on this thought, apart from the similarity measurement as them, computing formula is as follows by the Weighted H amming between the proper vector of calculating two width of cloth iris images for the present invention's proposition:
Here,
Iriscode
ABe the proper vector of input iris image,
Iriscode
i BRepresent i rotary template of registration iris image proper vector, because two width of cloth iris images cause the difference of translation, yardstick and rotation constantly owing to the variation of user's posture in different collections.By the pre-service of iris image, translation and dimensional variation are eliminated substantially, and we reduce the influence of rotation by the mode of the registration while mating,
WeightTemplate
iBe i rotary template of weighting template WeightTemplate (see figure 4), n is enrollment Iriscode
BTotal number of revolutions, generally be made as 7, the xor operation of symbol presentation logic position,
The following vector operations of symbol * be defined as:
(x
1，x
2，…，x
m)×(y
1，y
2，…，y
m)＝(x
1y
1，x
2y
2，…，x
my
m)，
Function S um is defined as
Weighted H amming distance H D
ABValue be floating number from 0 to 1, its value is more little, illustrates that the similarity of two width of cloth iris images is high more.
6. statistical learning weighting template S6, the weighting template is offline created in advance by the statistical learning of large sample, and each condition code is provided with different weights according to its independent discrimination, and discrimination is high more, and corresponding weights are also big more.Fig. 4 is the result of coupling in 3711 classes of study, and weights divide Three Estate to quantize, and can see that the weight in eyelashes and eyelid district is established lowlyer, meet people's visual impression.
7. recognition decision S7 is provided with different threshold values according to the different application scenario of iris authentication system, and corresponding different false acceptance rate (FAR, False Accept Rate) and false rejection rate (FRR, FalseReject Rate) are as Weighted H amming distance H D
ABDuring less than predefined threshold value, judge the user, otherwise provide not information by authenticating by authentication.
The robust features of above-mentioned iris image extracts and recognition methods, and its concrete steps are as follows:
1) mode conversion that the iris image under the cartesian coordinate system is used bilinear interpolation is under polar coordinates, and polar initial point is the center of circle of pupil, under polar coordinate system all iris images is zoomed to unified size, is called the normalization iris image;
2) with " multipole wave filter " the normalization iris image is carried out linear filtering: the multipole wave filter is roamed on the normalization iris image, and the gray-scale value of the image-region that is covered by the multipole wave filter and the window coefficient of wave filter carry out behind the dot product filtering result as the image pixel of corresponding multipole template center;
3) sign according to the filtering result is encoded into 0 or 1 (bit) with this image-region, and all binary codings are formed a proper vector, by bytes store;
4) calculate two width of cloth iris images (generally speaking, when being registration, gathers in advance a width of cloth iris image wherein, another width of cloth is to gather in the verification process temporarily) the Weighted H amming distance of feature coding, distance is more little, the similarity of two width of cloth images is high more, and is big more from the possibility of same eyes, and the weight of each bit iris feature sign indicating number correspondence depends on extensive training set the statistical recognition performance of condition code that should the zone iris image, discrimination is high more, and corresponding weights are also big more;
5) if Weighted H amming distance less than certain threshold value, then recognition system thinks that two width of cloth iris images come from same individual's same eyes, provides the information by authentication, otherwise prompting is by authentication.
Fig. 2 is an iris image pre-service synoptic diagram, wherein,
(a) be iris image;
(b) be the positioning result of (a); Wherein Bai Se circle is represented the outer boundary of pupil and iris after the match.
(c) be the normalization result of (b).
Fig. 3 is two examples of multipole wave filter, wherein,
(a) be one three utmost point subfilter, by two anodal sons (gauss low frequency filter of identical shaped and size, diverse location) and negative pole (Gaussian filter, window coefficient are-2 times of filter coefficient of forming anodal son) composition;
(b) be a quadrapole wave filter, by two anodal sons (gauss low frequency filter of identical shaped and size, diverse location) and two negative pole (Gaussian filter, window coefficient are-1 times of filter coefficient of forming anodal son) composition.
The weighting template of Fig. 4 for obtaining by off-line learning, in the polar coordinate system at normalization iris image place, the gray-scale value correspondence of weighting template the weights of the feature coding of relevant position, and the white portion weight is 1 among the figure, because these positions do not have blocking of eyelashes and eyelid substantially; The gray area weight is 0.5, is subjected to the influence that eyelashes and eyelid block because these positions are fewer; The black region weight is 0, because these positions often are subjected to the influence that eyelashes and eyelid block.
The distribution of the matching result that Fig. 5 tests on fairly large iris image database for the present invention, horizontal ordinate is that iris image mates the Weighted H amming distance that obtains in twos, ordinate is to be positioned at certain Hamming distance accounts for total matching times near matching result quantity ratio, from then on scheme we as can be seen between iris image from same eyes the Weighted H amming of (in the class) apart from the overwhelming majority less than 0.4, between iris image from different eyes the Weighted H amming of (between class) apart from the overwhelming majority greater than 0.4, so can determine very accurately that according to Weighted H amming distance the user is legal identity owner or personator.
The robust features that the present invention proposes a kind of iris image extracts and recognition methods.The advantage of this method is the following aspects:
1. accuracy of identification height.Because the multipole wave filter can be portrayed trickle and complicated image structure in the iris texture well, so the proper vector that obtains has very strong differentiation performance.Fig. 5 is the true iris image of 2155 width of cloth (totally 306 classes) that obtains according to the inventive method distribution plan of matching result in twos.It is overlapping to see that the distribution of mating between interior distribution of mating of class and class does not have substantially, and the property distinguished is good, has proved that the present invention has very high accuracy of identification.
2. code efficiency height, memory space is little.Owing to adopt binary-coding, the result of 8 filtering is as long as 1 byte just can be preserved, and the view picture iris image only needs the capacity storage of a hundreds of byte, can be kept at fully in the most IC-card (comprising second generation I.D.) and various mobile device.
3. computing velocity is fast.The main calculated amount consumption of whole algorithm is on the linear filtering of multipole wave filter and iris image, and this process can resolve into the low-pass filtering of several utmost points.If it is the process of low-pass filtering is approximate with fast algorithm, full feature is extracted and matching process only relates to addition and subtraction, there are not long multiplication and divide operations consuming time, and process is simple and clear, calculated amount is little, algorithm is easy to software programming and hardware is realized, the feature extraction time of testing on the common desk-top computer about 10 milliseconds, far below other classical iris identification method (generally about 50 milliseconds).
4. strong robustness.Low-pass filtering, non-neighbour comparison and binary-coding all make the robustness of whole algorithm strengthen greatly, and the influence that changed by noise, illumination contrast is little.
In sum, the present invention can finish iris recognition effectively, thereby carries out authentication reliably.Advantages such as the present invention simultaneously has that computing velocity is fast, accuracy of identification is high, strong robustness, memory space are little.The present invention has very wide range of application, and it can be used for ecommerce, E-Government, electronics military task and electronic police affair, and other need carry out the field that identity is differentiated.
Embodiment 1: the application of iris recognition in mobile stock system
The present invention is particularly suitable for calculating with storage resources platform in short supply and carries out iris recognition.Need speculate in the stock market with mobile phone as Mr. Li, he feels worried to carry out authentication with password, on his mobile phone iris authentication system is installed, take his iris image with digital camera after, though picture quality is not fine, point fuzziness is arranged, and the chip in the mobile phone still calculates the robust features coding (512 byte) of this width of cloth image very soon, and iris feature code encryption (and adding timestamp) back is sent to the stock exchange platform of short message receiving port as short message.Stock exchange trading system will obtain Mr.'s Li condition code after the short message deciphering, then with database in the condition code registered compare, matching distance is enough little.Stock exchange trading system is given the authority that Mr. Li operates personal stock informatino according to the result of identification, and notifies Mr. Li with the result at once.Mr. Li obtains identity authentication result after 3 seconds, safe Internet stock trading will begin in a minute.
Embodiment 2: the application of iris recognition in portable police system
The police finds that in the patrol process Zhang San's behavior is suspicious, start the iris authentication system among the police service PDA that carries at once, after taking Zhang San's iris image, automatically calculate it characteristics of image coding and with it with police system in up to ten thousand fugitive suspects' storing iris feature mate one by one, matching process was finished after 5 seconds, the coupling mark of discovery Zhang San's iris feature and a wanted criminal's iris is very high, and system sends warning message at once.Though Zhang San's profile has been passed through camouflage, this police has successfully caught the bad person under the help based on the iris authentication system of multipole wave filter.
Claims (13)
1. the robust features of an iris image extracts and recognition methods, comprises step:
The iris image pre-service;
Use " multipole wave filter " the normalization iris image is carried out airspace filter;
The result of filtering is carried out binary-coding, make up the proper vector of iris image;
Calculate the Weighted H amming distance between the proper vector of two width of cloth iris images;
According to the Weighted H amming that calculates gained apart from judging that two above-mentioned width of cloth iris images are whether from same people's same eyes.
2. method according to claim 1 is characterized in that, is operated under the two states: registration mode and recognition mode, and in registration mode, validated user is submitted the iris feature template of oneself to system; At recognition mode, system determines by the template of contrast storage and the feature of the interim iris image of gathering of user whether this user identity is legal, no matter be registration mode or recognition mode, iris identification method all will carry out image pre-service and feature extraction, also will carry out characteristic matching under recognition mode.
3. method according to claim 1, its step is as follows:
Iris image acquiring S1 just can photograph the iris image clearly that is rich in detail textures information with camera;
Iris image pre-service S2, the pretreated task of iris image is to find the coordinate and the radius of the circle of match pupil and iris boundary, then the iris annulus is normalized to the rectangular area of fixed measure, i.e. Iris Location and normalization;
Adopt " multipole wave filter " to carry out feature extraction S3, use the multipole wave filter normalization iris image of pre-service output is carried out airspace filter: the multipole wave filter is roamed on the normalization iris image, and the gray-scale value of the image-region that is covered by the multipole wave filter and the window coefficient of wave filter carry out behind the dot product filtering result as the image pixel of corresponding multipole template center;
Robust coding S4 carries out binary-coding to filtered image-region, the composition characteristic vector of at last all condition codes being joined together according to the filtering result's of previous step sign symbol;
Characteristic matching S5, when biological identification system operates under the certification mode, the user will state the identity of oneself, system finds out this people according to statement from database template characteristic vector sum input feature value mates;
Apart from the similarity measurement as them, computing formula is as follows by the Weighted H amming between the proper vector of calculating two width of cloth iris images:
Here,
Iriscode
ABe the proper vector of input iris image,
Iriscode
B iRepresent i rotary template of registration iris image proper vector, because two width of cloth iris images cause the difference of translation, yardstick and rotation constantly owing to the variation of user's posture in different collections, pre-service by iris image, translation and dimensional variation are eliminated substantially, reduce the influence of rotation by the mode of registration while mating
WeightTemplate
iBe i the rotary template of weighting template WeightTemplate,
N is enrollment Iriscode
BTotal number of revolutions, generally be made as 7;
Statistical learning weighting template S6, the weighting template is offline created in advance by the statistical learning of large sample, and each condition code is provided with different weights according to its independent discrimination, and discrimination is high more, and corresponding weights are also big more;
Recognition decision S7 is provided with different threshold values according to the different application scenario of iris authentication system, and corresponding different false acceptance rate (FAR, False Accept Rate) and false rejection rate (FRR, FalseReject Rate) are as Weighted H amming distance H D
ABDuring less than predefined threshold value, judge the user, otherwise provide not information by authenticating by authentication.
4. method according to claim 3 is characterized in that, iris image pre-service S2 comprises: Iris Location and normalization
I. Iris Location
The gray scale of human eye pupil is far below the peripheral region, isolate pupil region so can use threshold method, center of gravity that then should the zone is gone to the edge of match pupil as preliminary pupil center with variable-sized template near this point, best fitting result is exactly the result of pupil location;
II. normalization
In the mode of bilinear interpolation, the iris annulus of having good positioning can be carried out the rectangular area of spatial alternation to a fixed measure.
5. according to claim 1 or 2 or 3 described methods, it is characterized in that its concrete steps are as follows:
1) mode conversion that the iris image under the cartesian coordinate system is used bilinear interpolation is under polar coordinates, and polar initial point is the center of circle of pupil, under polar coordinate system all iris images is zoomed to unified size, is called the normalization iris image;
2) with " multipole wave filter " the normalization iris image is carried out linear filtering: the multipole wave filter is roamed on the normalization iris image, and the gray-scale value of the image-region that is covered by the multipole wave filter and the window coefficient of wave filter carry out behind the dot product filtering result as the image pixel of corresponding multipole template center;
3) sign according to the filtering result is encoded into 0 or 1 with this image-region, and all binary codings are formed a proper vector, by bytes store;
4) the Weighted H amming distance of the feature coding of calculating two width of cloth iris images, distance is more little, the similarity of two width of cloth images is high more, possibility from same eyes is big more, the weight of each bit iris feature sign indicating number correspondence depends on extensive training set the statistical recognition performance of condition code that should the zone iris image, discrimination is high more, and corresponding weights are also big more;
5) if Weighted H amming distance less than certain threshold value, then recognition system thinks that two width of cloth iris images come from same individual's same eyes, provides the information by authentication, otherwise prompting is by authentication.
6. method according to claim 1, it is characterized in that, described " multipole wave filter " differed by a plurality of positions, positive and negative differing, the low-pass filter that yardstick differs and shape differs constitutes jointly, each low-pass filter is called one " utmost point ", the coefficient summation of all wave filters is 0, promptly " multipole wave filter " is a kind of form differential filter flexibly, the Gabor wave filter, the first order derivative wave filter of Gaussian function, the second derivative wave filter of Gaussian function, general differential filters such as wavelet filter also belong to the special circumstances of " multipole wave filter ".
7. according to claim 1 or 4 described methods, it is characterized in that the low-pass filter of described composition " multipole wave filter " can be square or box-shaped wave filter, promptly to the average wave filter of filtering image zone summation.
8. method according to claim 1, it is characterized in that, to the airspace filter process of normalization iris image be exactly: the multipole wave filter is roamed on the normalization iris image, and the gray-scale value of the image-region that is covered by the multipole wave filter and the window coefficient of wave filter carry out behind the dot product filtering result as the image pixel of corresponding multipole template center.
9. according to claim 1 or 6 described methods, it is characterized in that, filtered image-region is carried out binary-coding, if the filtering result is greater than 0 then the feature coding of regional center pixel is 1 according to above-mentioned filtering result's sign symbol; If the filtering result is smaller or equal to 0 then the feature coding of regional center pixel is 0, all at last feature codings are composed in series the proper vector of this iris image, perhaps be expressed as binary matrix form, " multipole wave filter " but also quantification gradation coding of the result of filtering.
10. according to claim 1 or 4 or 5 or 6 or 7 described methods, it is characterized in that proper vector or the eigenmatrix of describing a width of cloth iris image can be proper vector or the eigenmatrixes that is constituted jointly by a plurality of different multipole wave filter coding results.
11. according to claim 1 or 7 or 8 described methods, it is characterized in that, can be down-sampled to proper vector to reduce the length of proper vector according to actual conditions, save storage space and match time.
12. method according to claim 1 is characterized in that, judges that whether two width of cloth iris images depend on from same people's same eyes whether their the Weighted H amming distance of proper vector is enough little.
13., it is characterized in that the weighting template is offline created in advance by the statistical learning of large sample according to claim 1 or 10 described methods, the weights of each condition code are decided according to its discrimination: discrimination is high more, corresponding weights are also big more.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200410081184 CN1760887A (en) | 2004-10-11 | 2004-10-11 | The robust features of iris image extracts and recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200410081184 CN1760887A (en) | 2004-10-11 | 2004-10-11 | The robust features of iris image extracts and recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1760887A true CN1760887A (en) | 2006-04-19 |
Family
ID=36706952
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200410081184 Pending CN1760887A (en) | 2004-10-11 | 2004-10-11 | The robust features of iris image extracts and recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1760887A (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101894256A (en) * | 2010-07-02 | 2010-11-24 | 西安理工大学 | Iris identification method based on odd-symmetric 2D Log-Gabor filter |
| CN101916361A (en) * | 2010-05-28 | 2010-12-15 | 深圳大学 | Iris feature design method and system based on 2D-DCT (2D-Discrete Cosine Transform) |
| CN101499129B (en) * | 2008-01-30 | 2010-12-22 | 中国科学院自动化研究所 | Distant range iris recognition system and method |
| CN101539990B (en) * | 2008-03-20 | 2011-05-11 | 中国科学院自动化研究所 | Method for selecting and rapidly comparing robust features of iris images |
| CN102209975A (en) * | 2008-10-08 | 2011-10-05 | 虹膜技术公司 | Method for acquiring region-of-interest and/or cognitive information from eye image |
| CN101615286B (en) * | 2008-06-25 | 2012-01-04 | 中国科学院自动化研究所 | Blind hidden information detection method based on analysis of image gray run-length histogram |
| CN102542243A (en) * | 2010-12-17 | 2012-07-04 | 北京理工大学 | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method |
| CN101556646B (en) * | 2009-05-20 | 2013-03-13 | 电子科技大学 | Iris sorting scheme based on kernel clustering |
| WO2013087028A1 (en) * | 2011-12-16 | 2013-06-20 | 北京天诚盛业科技有限公司 | Iris recognition method and iris recognition method based on multi-directional gabor and adaboost |
| CN103577813A (en) * | 2013-11-25 | 2014-02-12 | 中国科学院自动化研究所 | Information fusion method for heterogeneous iris recognition |
| CN103679141A (en) * | 2013-12-03 | 2014-03-26 | 浙江温医雷赛医用激光科技有限公司 | Rotary tracking algorithm for irises |
| CN104680128A (en) * | 2014-12-31 | 2015-06-03 | 北京释码大华科技有限公司 | Four-dimensional analysis-based biological feature recognition method and four-dimensional analysis-based biological feature recognition system |
| CN104778729A (en) * | 2014-01-09 | 2015-07-15 | 上海帝仪科技有限公司 | Iris extraction method and equipment under uneven illumination condition |
| CN106022782A (en) * | 2016-07-13 | 2016-10-12 | 孙智博 | Iris payment system |
| CN106803091A (en) * | 2015-11-25 | 2017-06-06 | 深圳怡化电脑股份有限公司 | A kind of recognition methods of note denomination and system |
| CN107315985A (en) * | 2016-04-26 | 2017-11-03 | 腾讯科技（深圳）有限公司 | A kind of iris identification method and terminal |
| CN107408195A (en) * | 2015-12-29 | 2017-11-28 | 厦门中控智慧信息技术有限公司 | A kind of iris identification method and device |
| CN109389094A (en) * | 2018-10-23 | 2019-02-26 | 北京无线电计量测试研究所 | A kind of stable iris feature extraction and matching process |
| CN110082072A (en) * | 2019-04-30 | 2019-08-02 | 中国科学院合肥物质科学研究院 | A kind of detector array target data processing method |
| CN110647796A (en) * | 2019-08-02 | 2020-01-03 | 中山市奥珀金属制品有限公司 | Iris identification method and device |
| CN110866507A (en) * | 2019-11-20 | 2020-03-06 | 北京工业大学 | Method for protecting mobile phone chatting content based on iris recognition |
| CN111666847A (en) * | 2020-05-26 | 2020-09-15 | 张彦龙 | Iris segmentation, feature extraction and matching method based on local 0-1 quantization technology |
| CN112101199A (en) * | 2020-09-15 | 2020-12-18 | 西安艾瑞生物识别科技有限公司 | Long-distance iris recognition method based on deep learning |
| CN114913638A (en) * | 2022-04-08 | 2022-08-16 | 湖北安源建设集团有限公司 | Fire-fighting access control management method and system based on Internet |
-
2004
- 2004-10-11 CN CN 200410081184 patent/CN1760887A/en active Pending
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101499129B (en) * | 2008-01-30 | 2010-12-22 | 中国科学院自动化研究所 | Distant range iris recognition system and method |
| CN101539990B (en) * | 2008-03-20 | 2011-05-11 | 中国科学院自动化研究所 | Method for selecting and rapidly comparing robust features of iris images |
| CN101615286B (en) * | 2008-06-25 | 2012-01-04 | 中国科学院自动化研究所 | Blind hidden information detection method based on analysis of image gray run-length histogram |
| CN102209975A (en) * | 2008-10-08 | 2011-10-05 | 虹膜技术公司 | Method for acquiring region-of-interest and/or cognitive information from eye image |
| CN101556646B (en) * | 2009-05-20 | 2013-03-13 | 电子科技大学 | Iris sorting scheme based on kernel clustering |
| CN101916361A (en) * | 2010-05-28 | 2010-12-15 | 深圳大学 | Iris feature design method and system based on 2D-DCT (2D-Discrete Cosine Transform) |
| CN101894256A (en) * | 2010-07-02 | 2010-11-24 | 西安理工大学 | Iris identification method based on odd-symmetric 2D Log-Gabor filter |
| CN102542243A (en) * | 2010-12-17 | 2012-07-04 | 北京理工大学 | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method |
| WO2013087028A1 (en) * | 2011-12-16 | 2013-06-20 | 北京天诚盛业科技有限公司 | Iris recognition method and iris recognition method based on multi-directional gabor and adaboost |
| CN103577813A (en) * | 2013-11-25 | 2014-02-12 | 中国科学院自动化研究所 | Information fusion method for heterogeneous iris recognition |
| CN103679141B (en) * | 2013-12-03 | 2016-10-05 | 浙江温医雷赛医用激光科技有限公司 | The rotation tracking of iris |
| CN103679141A (en) * | 2013-12-03 | 2014-03-26 | 浙江温医雷赛医用激光科技有限公司 | Rotary tracking algorithm for irises |
| CN104778729A (en) * | 2014-01-09 | 2015-07-15 | 上海帝仪科技有限公司 | Iris extraction method and equipment under uneven illumination condition |
| WO2015103953A1 (en) * | 2014-01-09 | 2015-07-16 | 上海帝仪科技有限公司 | Method and device for extracting iris image image under condition of non-uniform illumination |
| CN104680128A (en) * | 2014-12-31 | 2015-06-03 | 北京释码大华科技有限公司 | Four-dimensional analysis-based biological feature recognition method and four-dimensional analysis-based biological feature recognition system |
| CN106803091A (en) * | 2015-11-25 | 2017-06-06 | 深圳怡化电脑股份有限公司 | A kind of recognition methods of note denomination and system |
| CN107408195B (en) * | 2015-12-29 | 2020-06-23 | 厦门中控智慧信息技术有限公司 | Iris identification method and device |
| CN107408195A (en) * | 2015-12-29 | 2017-11-28 | 厦门中控智慧信息技术有限公司 | A kind of iris identification method and device |
| CN107315985A (en) * | 2016-04-26 | 2017-11-03 | 腾讯科技（深圳）有限公司 | A kind of iris identification method and terminal |
| CN107315985B (en) * | 2016-04-26 | 2020-12-04 | 腾讯科技（深圳）有限公司 | Iris identification method and terminal |
| CN106022782A (en) * | 2016-07-13 | 2016-10-12 | 孙智博 | Iris payment system |
| CN109389094A (en) * | 2018-10-23 | 2019-02-26 | 北京无线电计量测试研究所 | A kind of stable iris feature extraction and matching process |
| CN109389094B (en) * | 2018-10-23 | 2021-04-16 | 北京无线电计量测试研究所 | Stable iris feature extraction and matching method |
| CN110082072A (en) * | 2019-04-30 | 2019-08-02 | 中国科学院合肥物质科学研究院 | A kind of detector array target data processing method |
| CN110647796A (en) * | 2019-08-02 | 2020-01-03 | 中山市奥珀金属制品有限公司 | Iris identification method and device |
| CN110866507A (en) * | 2019-11-20 | 2020-03-06 | 北京工业大学 | Method for protecting mobile phone chatting content based on iris recognition |
| CN111666847A (en) * | 2020-05-26 | 2020-09-15 | 张彦龙 | Iris segmentation, feature extraction and matching method based on local 0-1 quantization technology |
| CN112101199A (en) * | 2020-09-15 | 2020-12-18 | 西安艾瑞生物识别科技有限公司 | Long-distance iris recognition method based on deep learning |
| CN114913638A (en) * | 2022-04-08 | 2022-08-16 | 湖北安源建设集团有限公司 | Fire-fighting access control management method and system based on Internet |
| CN114913638B (en) * | 2022-04-08 | 2024-03-01 | 湖北安源建设集团有限公司 | Fire control access control management method and system based on Internet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1760887A (en) | The robust features of iris image extracts and recognition methods | |
| CN101030244B (en) | Automatic identity discriminating method based on human-body physiological image sequencing estimating characteristic | |
| CN101923640B (en) | Method for distinguishing false iris images based on robust texture features and machine learning | |
| He et al. | Multi-patch convolution neural network for iris liveness detection | |
| Sheela et al. | Iris recognition methods-survey | |
| JP2018200716A (en) | System and method for biometric authentication in connection with camera-equipped devices | |
| Daouk et al. | Iris recognition | |
| Sanchez-Reillo et al. | Iris recognition with low template size | |
| Akhtar et al. | Face spoof attack recognition using discriminative image patches | |
| US20150178547A1 (en) | Apparatus and method for iris image analysis | |
| CN107169458B (en) | Data processing method, device and storage medium | |
| CN102902980B (en) | A kind of biometric image analysis based on linear programming model and recognition methods | |
| CN101833646A (en) | Living iris detection method | |
| Singh et al. | Iris recognition system using a canny edge detection and a circular hough transform | |
| CN108960088A (en) | The detection of facial living body characteristics, the recognition methods of specific environment | |
| CN108133187B (en) | The one-to-one iris identification method of dimensional variation invariant feature and the voting of more algorithms | |
| Rajalakshmi et al. | A multimodal architecture using Adapt‐HKFCT segmentation and feature‐based chaos integrated deep neural networks (Chaos‐DNN‐SPOA) for contactless biometricpalm vein recognition system | |
| Chen et al. | Iris recognition using 3D co-occurrence matrix | |
| Murugan et al. | Fragmented iris recognition system using BPNN | |
| CN102436591B (en) | Discrimination method of forged iris image | |
| CN114299569A (en) | Safe face authentication method based on eyeball motion | |
| Chaudhary et al. | A new template protection approach for Iris recognition | |
| Gad et al. | Iris recognition based on log-gabor and discrete cosine transform coding | |
| Cui et al. | An appearance-based method for iris detection | |
| Karthikeyan | Efficient Biometric IRIS Recognition System Using Fuzzy Neural Network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |