CN100414558C - Automatic fingerprint distinguishing system and method based on template learning - Google Patents
Automatic fingerprint distinguishing system and method based on template learning Download PDFInfo
- Publication number
- CN100414558C CN100414558C CNB021545219A CN02154521A CN100414558C CN 100414558 C CN100414558 C CN 100414558C CN B021545219 A CNB021545219 A CN B021545219A CN 02154521 A CN02154521 A CN 02154521A CN 100414558 C CN100414558 C CN 100414558C
- Authority
- CN
- China
- Prior art keywords
- template
- fingerprint
- class
- fingerprint image
- 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.)
- Expired - Lifetime
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 239000000284 extract Substances 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims description 26
- 230000008878 coupling Effects 0.000 claims description 24
- 238000010168 coupling process Methods 0.000 claims description 24
- 238000005859 coupling reaction Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 7
- 238000013519 translation Methods 0.000 claims description 7
- 238000007621 cluster analysis Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 5
- 230000008439 repair process Effects 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 16
- 239000004744 fabric Substances 0.000 description 14
- 238000000605 extraction Methods 0.000 description 9
- 238000013461 design Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000015572 biosynthetic process Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000003909 pattern recognition Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- AYFVYJQAPQTCCC-GBXIJSLDSA-N L-threonine Chemical compound C[C@@H](O)[C@H](N)C(O)=O AYFVYJQAPQTCCC-GBXIJSLDSA-N 0.000 description 2
- 235000012364 Peperomia pellucida Nutrition 0.000 description 2
- 240000007711 Peperomia pellucida Species 0.000 description 2
- 230000033228 biological regulation Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000009897 systematic effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 239000010749 BS 2869 Class C1 Substances 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Landscapes
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to an automatic fingerprint recognition method based on template learning. The present invention comprises the steps of fingerprint image register, namely recording characteristic information of a fingerprint template and a fingerprint image to be matched; fingerprint image recognition, namely judging whether the characteristic information of the input fingerprint image is similar to some template information stored in a system database; detail information feedback, namely utilizing knowledge rules to feeding back the matched fingerprint image information to complete reliable detail point classes to which multiple templates correspond. In the present invention, the detail points of a plurality of templates of the same fingerprints are classified, the cores of the classes and the information of inter-class distance, class-in distance, etc. are stored, and a link of coarse matching is added; of the coarse matching fails, the ordinary matching operation is not carried out. The present invention can accurately extract and repair the characteristic information of multiple plates, enhances the recognition effect and performance of an automatic fingerprint recognition system and has an important application value in biologic recognition technology.
Description
Technical field
The invention belongs to the living things feature recognition field, particularly utilize cluster analysis and realize the matching process of multi-template fingerprint image based on the method for knowledge.
Background technology
Nineteen nineties, fingerprint identification technology becomes a kind of living things feature recognition method of maturation gradually, and it belongs to " pattern-recognition " field.At first, by the algorithm for recognizing fingerprint of a series of complexity, just can in the extremely short time, finish your identity identifying and authenticating then according to fingerprint with the fingerprint input computing machine that extracts.At present, Automated Fingerprint Identification System is widely used in the field that needs identity authentication.Automatically the application of fingerprint recognition no longer only is confined to law, public security field, it can be used as the means that computing machine is confirmed the user, can be used as the information security technology of accesses network resource, also can be used for many aspects such as double acknowledge, employee's proof and domestic electronic door lock of the affirmation of bank ATM card and credit card use, all kinds of intellective IC cards.Along with Automated Fingerprint Identification System the gate inhibition, work attendance, the widespread use of civil areas such as social security, people have also proposed more and more higher requirement to the accuracy of fingerprint recognition.
The formation of typical case's Automated Fingerprint Identification System as shown in Figure 1.
Typical case's Automated Fingerprint Identification System mainly comprises several big modules (A.K.Jain and S.Pankanti such as fingerprint collecting, feature extraction and coupling, " Automated Fingerprint Identification andImaging Systems ", Advances in Fingerprint Technology, 2nd Ed. (H.C.Leeand R.E.Gaensslen), CRC Press, 2001.).From the angle of modern scientific research, the recognition methods of this type systematic relates generally to problems such as fingerprint image acquisition, fingerprint image enhancing, feature extraction, preservation data, characteristic matching.Wherein feature extraction and fingerprint matching are two key problems of fingerprint recognition system, also are two basic problems and the important research project of pattern-recognition.
According to the standard of U.S. FBI, usually fingerprint is divided into crestal line and valley line, think that the end of crestal line puts slightly with bifurcation and have lifelong unchangeability and uniqueness, as Fig. 2.
Therefore, algorithm for recognizing fingerprint commonly used is the end of fingerprint image to be put slightly with bifurcation extract as characteristic information, is rotated between the fingerprint characteristic point set then after the calibration of translation, calculates the process of similarity by unique point.The authoritative scholar A.K.Jain in international living things feature recognition field leads Univ Michigan-Ann Arbor USA Flame Image Process and pattern-recognition seminar these to be carried out deep research, and has applied for multinomial U.S. patent of invention, and for example: U.S.Patent 6,185,318, Feb.6,2001.
This method that they proposed is had relatively high expectations to the Template Information of fingerprint, and the details in fingerprint that it is comprised in will seeking template is counted out can not be very little, and information such as locality must be accurately.When the fingerprint template useful area of gathering hour, the details of fingerprint template is counted, and less (for example: the Acquisition Instrument of Authentic company can only collect the image of 128 * 128 pixels, the fingerprint image of gathering can only be the part of actual fingerprint, and it is few and be difficult to coupling that such image extracts minutiae point); When fingerprint image was second-rate, the minutiae point possible position direction of extraction etc. had fake minutiae to exist.Under above two kinds of situations, the method that they adopted easily produces the phenomenon that mistake is known or refused to know so.Therefore, in Automated Fingerprint Identification System, according to the finger print information on the coupling, removing the also additional fingerprint characteristic more accurately of fake minutiae and repair fingerprint template, is a kind of effective way that improves recognition effect, also is that problem to be solved is arranged.
Summary of the invention
The objective of the invention is to propose and design a kind of automatic fingerprint identification method and system of practicality.Can handle many pieces of template images of same fingerprint, the common information of many pieces of templates is fused into whole multi-template details category information, then this kind of information is applied to fingerprint Matching Algorithm; Can in use feed back the information of the fingerprint image on the coupling simultaneously, be used for the parameter of learning training multi-template details category information.Make Template Information abundant accurately, reduce the template individual difference and refusing of causing known and mistake is known phenomenon.
For achieving the above object, according to an aspect of of the present present invention, comprise step based on the automatic fingerprint identification method of Template Learning:
(1) based on the registering fingerprint of multi-template detailed information cluster, write down the characteristic information of fingerprint template and fingerprint image to be matched, comprising:
Several template image calibrations;
Multi-template detailed information cluster calculates between class distance in the nuclear of many pieces of template corresponding reliable detail points and the class;
(2) based on the fingerprint image identification of detailed information feedback, utilize knowledge rule, the information in fingerprint on the feedback coupling judges whether the characteristic information of input fingerprint image is similar with certain the fingerprint template information in the system database, comprising:
Fingerprint image slightly mates, and judges the minutia point set of fingerprint image and the similarity degree between multi-template minutiae point class;
Fingerprint image carefully mates, the output matching result;
The detailed information feedback.
According to another aspect of the present invention, a kind of Automated Fingerprint Identification System based on Template Learning comprises:
(1) based on the registering fingerprint module of multi-template detailed information cluster, write down the characteristic information of fingerprint template and fingerprint image to be matched, comprising:
Several template image calibrations;
Multi-template detailed information cluster calculates many pieces of templates between class distance in the nuclear of the reliable detail points of English and the class;
(2) based on the fingerprint image identification module of detailed information feedback, utilize knowledge rule, the information in fingerprint on the feedback coupling judges whether the characteristic information of input fingerprint image is similar with certain the fingerprint template information in the system database, comprising:
Finger print image slightly mates, and judges the minutia point set of fingerprint image and the similarity degree between multi-template minutiae point class;
Fingerprint image carefully mates, the output matching result;
The detailed information feedback.
The present invention will be with the classification of the minutiae point of a plurality of templates of fingerprint, has write down between the nuclear of these classes and class in the class information such as distance, has increased the link of thick coupling, if thick coupling gets nowhere, then no longer carries out general matching operation.The present invention can accurately extract and repair the characteristic information of multi-template, improves the recognition effect and the performance of Automated Fingerprint Identification System, has important use to be worth in biological identification technology.
Description of drawings
Fig. 1 is the formation of typical Automated Fingerprint Identification System;
Fig. 2 is the fingerprint feature;
Fig. 3 is the Automated Fingerprint Identification System synoptic diagram of band feedback
Fig. 4 is the fingerprint image that same finger is repeatedly gathered;
Fig. 5 is based on the formation of the Automated Fingerprint Identification System of Template Learning;
Fig. 6 is the enhancement process flow process of fingerprint image;
Fig. 7 is the 8 connection neighborhood presentation graphs that M is ordered;
Fig. 8 is the minutiae point model;
Fig. 9 is minutiae point class figure and nuclear feature;
Figure 10 is several possible results in the cluster process, and wherein, figure a is the similar point set in the same class; Figure b is the dissimilar point set of angle; Figure c is an isolated point; Figure d be similitude to but be in image border or out-of-bounds
Figure 11 is the fingerprint image recognition system that designs realization voluntarily;
Figure 12 is that wherein, test figure is a fingerprint image with the generation of finger multi-template, and resolution is 300 * 300 * 256.Figure a, figure b and figure c are respectively the template images that same finger is gathered for three times; Figure d is the calibration image of figure a, figure b and figure c; Figure e is the nuclear collection of figure d multi-template cluster;
Figure 13 is that identifying is given an example;
Figure 14 is a matching result statistics ROC curve map;
Figure 15 is FVC evaluation of algorithm result.
Embodiment
Our method proposes the new model of template representation, the finger print information that the match is successful can be used for repairing improving fingerprint template, promptly increases feedback element in traditional Automated Fingerprint Identification System, as Fig. 3.Simultaneously, our fingerprint image recognition methods can be removed the fake minutiae that the original fingerprint image obtains because of noise effect, writes down the characteristic information of fingerprint as far as possible exactly, can both debate to know for fingerprint image of poor quality or that minutiae point is few and handle.Therefore, our fingerprint image recognition methods has adapted to the novel finger print Acquisition Instrument and has been tending towards the requirement that miniaturization develops, and has guaranteed the robustness of Automated Fingerprint Identification System.
Automated Fingerprint Identification System should be able to carry out real-time coupling and debate the knowledge operation jumbo fingerprint database efficiently.Usual way is that the input fingerprint is classified earlier by types such as bucket type, left whirlpool type, right whirlpool type, arch form, cusped arch types, and the fingerprint base in its place class carries out man-to-man coupling then.Problem hereto, our method resolution policy is: according to special template representation method, designed thick coupling and two steps of thin coupling.Have only template that slightly the match is successful and the details point set of importing fingerprint to carry out man-to-man thin coupling in the template base, have only the template that carefully the match is successful just to participate in parameter learning and reparation.Such operation makes our method have high efficiency and accuracy.
Because fingerprint acquisition instrument and gathered a variety of causes such as people, the finger-print region that collects of fingerprint sensor is smaller sometimes, and such fingerprint can not provide sufficient information for the automatic system of fingerprints of high discrimination, such as minutiae point.And, sometimes with same finger collection to fingerprint image also may have only the sub-fraction zone overlapped, so also can have influence on the matching performance of fingerprint recognition system.Therefore, the fingerprint with the multiple angles of same finger continuous acquisition in registration process all saves as template, and this is a kind of measure that improves the matching rate of fingerprint.As Fig. 4, be the fingerprint image that same finger is repeatedly gathered.
Yet, the fingerprint of each width of cloth registration all independently as template, is not considered correlativity and common trait between these templates, such way will cause the redundant information magnanimity of template database to increase and the waste of resource.How to extract the public information of many pieces of fingerprint templates, the fake information of removing indivedual templates is the problem that needs solution in order to improve the system identification performance.Our method adopts the method for cluster analyses such as improved Clique figure to merge the information of a plurality of fingerprint templates, is used for the identification of fingerprint, and the result of implementation illustration method is practical reliable.
Core concept of the present invention is to take effective clustering method and simulate the way of manually doing fingerprint image identification based on the training study method of knowledge and with computing machine.Because fingerprint image has it self, and two main prioris that can be used to carry out identification and matching are arranged, the one, the distribution of fingerprint minutiae point position, the 2nd, near the texture the fingerprint minutiae point.Near the grain direction of the distribution of fingerprint minutiae position and corresponding fingerprint minutiae is similar between a plurality of templates of same fingerprint, and for the image of different finger collections, is diverse.We can conclude the general character of these visual informations between a plurality of templates of summing up same fingerprint, and the minutiae point of fingerprint are distributed such structural information shows in computing machine, in the identifying afterwards, can distinguish the similarities and differences of fingerprint exactly again.The people is introduced the understanding of dactylotype in the process of fingerprint image coupling, with computing machine simulate the way of manually doing images match be necessary also be possible.This image matching algorithm utilizes the structural information of fingerprint image to come the process of navigational figure coupling based on people to the understanding (being two main prioris of fingerprint image) of dactylotype with the form of rule just.
Describe in detail based on the fingerprint image matching algorithm of Template Learning and the design of recognition system below.As concrete recognition system, main modular has: registering fingerprint module, fingerprint image identification module and feedback module.For wherein concrete recognizer, key step is respectively: minutiae point is extracted, multi-template detailed information cluster, and fingerprint image slightly mates, and fingerprint image carefully mates, and the multi-template detailed information is repaired.Below it is made introductions all round.
As shown in Figure 5, the module of system mainly is divided into the registering fingerprint module, fingerprint image identification module and feedback module.
Registering fingerprint is meant at off-line and gathers in the process of fingerprint that fingerprint image is as template (usually, 3≤N≤10) preferably to need several quality of each fingerprint collecting, and the minutia information of extraction and record fingerprint is stored in the template database.
The main processing procedure of this module has: image acquisition, Flame Image Process and minutiae point are extracted
Step 1: the collection of fingerprint image
Acquisition method has printing ink to push and two kinds of instrument collections.Can select a certain optical sensor, CMOS fingerprint sensor for use, heat sensitive sensor, novel sensors such as ultrasonic sensor are as the equipment of fingerprint collecting.Require as far as possible the zone level at place, fingerprint singularity center to be placed on the center of acquisition chip, by being pressed with certain dynamics.
Need each fingerprint collecting N width of cloth quality preferably fingerprint image as template, usually, 3≤N≤10.
Step 2: the enhancement process of fingerprint image
The enhancement process of fingerprint image refers to the process of using some image processing meanses that fingerprint image is processed.In our fingerprint algorithm, this step is relatively more crucial.Treatment scheme as shown in Figure 6.
Concrete processing operation has: 1. the equalization of gray scale, this can eliminate the difference of contrast between the different images.2. use simple low-pass filtering algorithm to eliminate speckle noise and Gaussian noise.3. calculate the border of image, carry out the cutting of image.Can reduce next step amount of calculation like this, improve the speed of system.4. the estimation of the field of direction calculates the direction of each pixel of fingerprint image.5. binaryzation comes fingerprint image is treated to the image that has only black and white two looks according to the direction of each picture element.6. refinement according to the image of binaryzation, to having only a pixel, generates fingerprint thinning figure to the crestal line width reduction of fingerprint.7. some tangible broken strings in the refined image are removed in refinement aftertreatment, bad crestal line structures such as the burr between crestal line on tangible bridge, the crestal line, too short crestal line and single spot.
Step 3: detail extraction
We use following algorithm to detect minutiae point: as shown in Figure 7,
The M that sets up an office represents the gray-scale value on the refined image, and M=0 represents that this point is stain, and M=255 is expressed as white point.
If M=0, and
Then M is a destination node;
If M=0, and
Then M is a bifurcation.
Because the recorded information of minutiae point is to determine according to concrete matching algorithm.Fig. 8 is the minutiae point model of our method.We write down following information according to our matching algorithm:
1) x of minutiae point, the y coordinate
2) the direction θ of minutiae point, this direction is defined as the direction of the local crestal line at this minutiae point place.
3) the type t of minutiae point, i.e. crestal line tip or crestal line branch.
So just a width of cloth fingerprint image has been changed into a plane point set M={M who forms by minutiae point
k, 1≤k≤L}.Wherein L is the number of the concentrated minutiae point of point.For any one minutiae point wherein, its eigenvector is
The fingerprint image identification module is exactly that input fingerprint image to be identified and the Template Information in the system database are mated, so judge the input fingerprint whether with template base in certain piece of fingerprint from same finger.This module can be divided into off-line and online two parts, wherein: in the off-line part, to after the calibration of fingerprint N width of cloth template, characteristic is carried out cluster analysis, data such as the nuclear of calculated characteristics point; In online part, after the details point set calibration to the input fingerprint image, do matching operation with the data of template base.This two-part method of operating is described respectively below.
1. off-line part
Step 1: from the N width of cloth template image calibration of same finger
Because the influence of the factors such as time environment of fingerprint collecting, even same fingerprint collecting to several fingerprint images can not overlap fully, and can rotate and translation.Must be before doing template cluster and fingerprint matching different fingerprint image calibrations.Respectively with the 2nd to N width of cloth template point set M of each finger in the fingerprint base
I, j T(j=2 ..., n), with the 1st width of cloth template point set M of correspondence
I, 1 TFor benchmark is rotated and translation transformation.Concrete, the details point set of a fingerprint image is as follows to the details point set Calibration Method of another fingerprint image:
If the details point set of two width of cloth fingerprint images is
Wherein point set P is total to M point, and point set Q is N point altogether.I some p for point set P
i(1≤i≤M), (p
x i, p
y i) be the x and the y axial coordinate of minutiae point,
Be the direction of minutiae point, t
P iType for minutiae point; J some q for point set Q
j(1≤j≤N), (q
x j, q
y j) be the x and the y axial coordinate of minutiae point,
Be the direction of minutiae point, t
j QType for minutiae point.
Purpose is to seek optimal mapping F
S, θ, Δ x, Δ y: R
2→ R
2,
Make F
S, θ, Δ x, Δ y(p)=q.Here Δ θ is a rotation parameter, and (Δ x, Δ y) is translation parameters, and they belong to the attitude calibration parameter;
Be the reference minutiae point.
Our method is that search two width of cloth details point set k (recommendation k=5) the most similar individual point is right, then respectively with each point to reference point the most, the rotation parameter and the translation parameters of the local detail point set of estimating according to the method for A.K.Jain.Be limited set just with each parameter discrete:
Δθ∈{Δθ
1，Δθ
2，...Δθ
L}，Δx∈{Δx
1，Δx
2，...Δx
L}，Δy∈{Δy
1，Δy
2，...Δy
L}
Wherein, the right computing method of similitude of two width of cloth details point set k are similar to Xudong Jiang (Xudong Jiang, Wei-Yun Yau.Fingerprint Minutiae Matching Based on theLocal and Global Structures.ICPR 2000:6038-6041) method, calculate promptly that a little (* *) gets the preceding k of maximum value to point to similarity function sl.
Respectively with the 2nd to N width of cloth template point set M of each finger in the fingerprint base
I, j T(j=2 ..., n), with the 1st width of cloth template point set M of correspondence
I, 1 TFor benchmark be rotated with translation transformation after, the regulation mould plate point set that obtains is designated as M
I, j A, T (j=1 ..., N), wherein
(i＝1，…，L
1)。
Step 2: multi-template detailed information cluster
The minutia that it has been generally acknowledged that fingerprint has uniqueness and unchangeability throughout one's life, is certainly existing the corresponding mutually similarity relation of some minutiae point between the details point set that obtains respectively after treatment through the template of repeatedly gathering so.The true minutiae point of a template point set should be concentrated at other template points respectively and all find similar some correspondence; And the fake minutiae of this template point set is concentrated at other template points and is not had corresponding point.We adopt the thought of cluster analysis, and minutiae point similar between the template point set is included in the class, and dissimilar minutiae point is all outside class.Be similar similarity maximum, inhomogeneity similarity minimum.We define the general character that the nucleoid proper vector is described similar minutiae point in the class then.Concrete grammar is as follows:
Input: the regulation mould plate point set is designated as M
I, j A, T(j=1 ..., N)
Output: minutiae point info class C
i={ C
1..., C
lAnd the nuclear proper vector of class
So after the calibration through the front, the 2nd to N width of cloth template point set M
I, j T(j=2 ..., n) in the 1st width of cloth template point set M
I, 1 TCorresponding some p of certain minutiae point q
j(j=2 ..., n) will drop on the close region of a q, be minutiae point class figure and nuclear feature as Fig. 9.
So we to define similarity function as follows:
If M
I, 1(1≤i≤L
1) be template point set M
1Minutiae point, M
J, 2(1≤j≤L
2) be template point set M
2Minutiae point, we define M
I, 1And M
J, 2Be similar, if satisfy condition
(1) | M
I, 1-M
J, 2|＜Thre and
(2) there is M ' respectively
I, 1And M "
I, 1Be M
I, 1At its template point set M
1In two neighbours, have M ' respectively
J, 2And M "
J, 2Be at its template point set M
2In two neighbours. satisfy
| M '
I, 1-M '
J, 2|＜T and | M "
I, 1-M "
J, 2|＜T
Here, Thre and T are threshold parameters.
As Figure 10, classifying rules is as follows:
1. if the similar r ∈ that counts [4, N] is designated as credible class with these similitudes
2. if the similar r ∈ [2,3] and be positioned at the image border of counting is designated as candidate's class with these similitudes
3. other isolated discrete point mark not is not designated as class yet.
Our point that just will satisfy the similarity condition is classified as a class like this.Under regard to each class C, we describe aggregation extent and the partial structurtes features such as mean direction and position thereof between similitude in the class with a proper vector.We define the nuclear proper vector K of this proper vector for such.
The nuclear proper vector computing method of class are as follows:
If the class of template similitude is C
i={ M
I, j, j=1 ..., L}, here
Vector
Be designated as i class C
iThe nuclear vector, here,
Minutiae point average coordinates in the class:
Minutiae point mean direction in the class:
The boundary box radius of class:
(λ is the parameter greater than 1)
2. online part
Step 1: fingerprint image slightly mates.If success, the template numbering that record may mate forwards following step 2 to; If the matching result unsuccessful, that output " does not have the coupling fingerprint ".
Concrete thick matching process is as follows:
After the foregoing enhancement process of fingerprint image process of step (1) with input, extract minutiae point vector set M
I
Step (2) the nuclear set of eigenvectors K of each template of template base
TCompare with it, judgement input picture minutiae point colony is gone into the R that counts of the r radius region of template cluster nuclear,
Step (3) number R thinks then that when greater than given threshold value slightly the match is successful.Illustrate this template may with the input images match.Add this template label that may mate to candidate list, forward thin matching operation (being step 2) to.Otherwise, forward step (2) to, search next template, all templates traversal once withdraws from the storehouse, returns that it fails to match.
Step 2: fingerprint image carefully mates.Mate one by one with possible fingerprint template, if the template of coupling is arranged, the matching result of output " the match is successful "; If the matching result unsuccessful, that output " does not have the coupling fingerprint ".
Concrete thin matching process is as follows:
The minutiae point vector set M of step (1) calibration input fingerprint
IDetails point set M with candidate template
1 T
Step (2) adopts Xudong Jiang (Xudong Jiang, Wei-Yun Yau.FingerprintMinutiae Matching Based on the Local and Global Structures.ICPR 2000:6038-6041) method is carried out polar coordinate transform, and calculates the coupling mark
Step (3) is if the coupling mark greater than given threshold value, thinks that then carefully the match is successful, returns M
IBe used for systematic learning.Otherwise, search next candidate template, return step (1), up to all candidate template traversal once, withdraw from, return that it fails to match.
The purpose of feedback module is to utilize the template data of the input information in fingerprint reparation fingerprint on the coupling, and repetition training and study through to template data more meet it.Its process flow diagram is as figure.If the details point set of input fingerprint image is M
I, know certain fingerprint matching in it and the database, the template set M of this fingerprint through debating
T
Specifically mainly contain 4 steps.
Step 1: with M
IWith M
TMerge into a new minutiae point set M.
Step 2: the cluster result C that calculates new minutiae point set M with the cluster analysis device
i *
Step 3: calculate cluster result C
i *In average similarity H between similar minutiae point
Ave *And the weighted mean similarity S between class
Ave *, same, for M
TCluster result C
i, calculate H
AveAnd S
AveMethod is as follows:
Wherein, F (M
i) and F (K) be respectively minutiae point M
iEigenvector with centronucleus K.N is the number of minutiae point in the class.(* *) is the similarity discriminant function of eigenvector to S.
For the class C1 of minutiae point clustering operation back formation ..., Ct,
Step 4: if satisfy condition:
And
Revise the Template Information of corresponding fingerprint, with C
i *Replaced C
i, recomputate the nuclear and the class radius of class and each class of minutiae point.
Show that through test this Fingerprint Image Recognition Algorithms can be to the compression of classifying of the information of a plurality of templates of same finger, the common details feature point set of extraction is accurate, in system increase slightly mate and feed back two links after effect very good.In the process of fingerprint recognition, can be good at using.
Embodiment
As shown in Figure 1, we design the fingerprint image recognition system of realization voluntarily.
Fingerprint image processing system is based on Window98/95, adopts Object Oriented method and soft project standard, Flame Image Process and analytic system that realize with C Plus Plus, object fingerprint identification field.Native system has abundant graph and image processing and analytic function, not only has perfect two dimensional image Treatment Analysis function, and can the various algorithm for recognizing fingerprint of dynamic load.System provides the image input, the image storage, and Flame Image Process, algorithm loads, file conversion, a series of functions such as FVC testing tool.
Below to specific implementation process based on the automatic fingerprint identification method of Template Learning.Test figure is the database of FVC2000, and resolution is 300 * 300 * 256.
1) reads in several fingerprint template images by opening file or opening button.
2) click the load-on module menu and load the fingerprint enhancement algorithms.
3) click the extraction minutiae point, obtain smooth fingerprint thinning figure, comprise tip point and bifurcation, as Figure 12.
4) calibrate the feature point set of several fingerprint template images, as Figure 12 (d).
5) generate template data, as Figure 12 (e).
6) read in single width input fingerprint image by opening file or gathering the fingerprint button.
7) respectively with fingerprint enhancing and extraction details operational processes fingerprint to be identified, as Figure 13 (b).
8) choose file in fingerprint to be identified and the template base.
Method adopts the recognizer criterion evaluation method of international fingerprint recognition contest and the fingerprint database FVC2000 of standard to test, experimental result such as Figure 14 and shown in Figure 15,
The above results is consistent to the theoretical analysis conclusion of research of fingerprint image matching algorithm and system design with the inventor.Has high reliability, applicability and admissibility.
Claims (7)
1. automatic fingerprint identification method based on Template Learning comprises step:
(1) based on the registering fingerprint of multi-template detailed information cluster, write down the characteristic information of fingerprint template and fingerprint image to be matched, comprising:
Several template image calibrations;
Multi-template detailed information cluster calculates between class distance in the nuclear of many pieces of template corresponding reliable detail points and the class;
(2) based on the fingerprint image identification of detailed information feedback, utilize knowledge rule, the information in fingerprint on the feedback coupling judges whether the characteristic information of input fingerprint image is similar with certain the fingerprint template information in the system database, comprising:
Fingerprint image slightly mates, and judges the minutia point set of fingerprint image and the similarity degree between multi-template minutiae point class;
Fingerprint image carefully mates, the output matching result;
The detailed information feedback.
2. by the described method of claim 1, it is characterized in that described several template image calibrations comprise step:
The most similar a plurality of points are right between the calculation template point set;
Estimate every pair of right rotation parameter and translation parameters of similitude;
At the right regional area calibration details point set of similitude.
3. by the described method of claim 1, it is characterized in that described multi-template detailed information cluster comprises step:
Similarity function by definition is judged similar point set;
Classifying rules by definition is divided class.
4. by the described method of claim 1, it is characterized in that described multi-template detailed information cluster comprises recording step:
Minutiae point average coordinates (x in the class
c, y
c);
Minutiae point mean direction in the class
The boundary box radius of class
5. by the described method of claim 1, it is characterized in that the thick coupling of described finger print image comprises step:
After the finger print image process enhancement process with input, extract minutiae point vector set M
1
Nuclear set of eigenvectors K with each template of template base
TCompare with it, judgement input picture minutiae point colony is gone into the R that counts of the r radius region of template cluster nuclear,
If number R thinks then that when greater than given threshold value slightly the match is successful, add this template label that may mate to candidate list, forward thin matching operation to, otherwise, forward step (2) to, search next template, all templates traversal once returns that it fails to match in the storehouse.
6. by the described method of claim 1, it is characterized in that described detailed information feedback comprises step:
Point set is asked also, and input fingerprint image characteristics point set and corresponding templates feature point set merge;
Cluster is calculated the cluster result of new minutiae point set with the cluster analysis device;
Average similarity in the compute classes attribute, class and the average similarity between class;
Revise template, recomputate the nuclear proper vector of class.
7. Automated Fingerprint Identification System based on Template Learning comprises:
(1) based on the registering fingerprint module of multi-template detailed information cluster, write down the characteristic information of fingerprint template and fingerprint image to be matched, comprising:
Several template image calibrations;
Multi-template detailed information cluster calculates many pieces of templates between class distance in the nuclear of the reliable detail points of English and the class;
(2) based on the fingerprint image identification module of detailed information feedback, utilize knowledge rule, the information in fingerprint on the feedback coupling judges whether the characteristic information of input fingerprint image is similar with certain the fingerprint template information in the system database, comprising:
Finger print image slightly mates, and judges the minutia point set of fingerprint image and the similarity degree between multi-template minutiae point class;
Fingerprint image carefully mates, the output matching result;
The detailed information feedback.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB021545219A CN100414558C (en) | 2002-12-06 | 2002-12-06 | Automatic fingerprint distinguishing system and method based on template learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB021545219A CN100414558C (en) | 2002-12-06 | 2002-12-06 | Automatic fingerprint distinguishing system and method based on template learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1506903A CN1506903A (en) | 2004-06-23 |
| CN100414558C true CN100414558C (en) | 2008-08-27 |
Family
ID=34235516
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB021545219A Expired - Lifetime CN100414558C (en) | 2002-12-06 | 2002-12-06 | Automatic fingerprint distinguishing system and method based on template learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100414558C (en) |
Families Citing this family (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7813531B2 (en) * | 2006-05-01 | 2010-10-12 | Unisys Corporation | Methods and apparatus for clustering templates in non-metric similarity spaces |
| CN101123500B (en) * | 2006-08-11 | 2011-02-02 | 华为技术有限公司 | A biologic verification method and device |
| CN101165704B (en) * | 2006-10-20 | 2010-07-14 | 西安紫牛信息技术有限公司 | Composite fingerprint template matching method |
| CN101751550B (en) * | 2008-12-19 | 2012-02-01 | 杭州中正生物认证技术有限公司 | Fast fingerprint searching method and fast fingerprint searching system thereof |
| CN101777128B (en) * | 2009-11-25 | 2012-05-30 | 中国科学院自动化研究所 | Fingerprint minutiae matching method syncretized to global information and system thereof |
| CN101777115B (en) * | 2009-11-25 | 2012-02-15 | 中国科学院自动化研究所 | Safe fingerprint verification method and system |
| CN102136024B (en) * | 2010-01-27 | 2013-01-02 | 中国科学院自动化研究所 | Biometric feature identification performance assessment and diagnosis optimizing system |
| CN102034114A (en) * | 2010-12-03 | 2011-04-27 | 天津工业大学 | Characteristic point detection-based template matching tracing method |
| JP5792320B2 (en) * | 2010-12-29 | 2015-10-07 | トムソン ライセンシングＴｈｏｍｓｏｎ Ｌｉｃｅｎｓｉｎｇ | Face registration method |
| WO2012141287A1 (en) * | 2011-04-15 | 2012-10-18 | 株式会社エヌ・ティ・ティ・ドコモ | Portable terminal, and grip characteristic learning method |
| US20130214905A1 (en) * | 2011-04-15 | 2013-08-22 | Ntt Docomo, Inc. | Portable terminal and gripping-feature learning method |
| US20140003681A1 (en) * | 2012-06-29 | 2014-01-02 | Apple Inc. | Zero Enrollment |
| CN103020590B (en) * | 2012-11-20 | 2015-12-09 | 北京航空航天大学深圳研究院 | A kind of vehicle identification system based on three-dimensional model and images match and method thereof |
| CN103077377B (en) * | 2012-12-31 | 2015-07-29 | 清华大学 | Based on the fingerprint correction method of field of direction distribution |
| CN103177240B (en) * | 2013-02-05 | 2016-04-27 | 金硕澳门离岸商业服务有限公司 | General-purpose fingerprint template generation device and method |
| CN103679201B (en) * | 2013-12-14 | 2017-01-11 | 复旦大学 | Calibration method of point set matching for image matching, recognition and retrieval |
| WO2015104115A1 (en) * | 2014-01-07 | 2015-07-16 | Precise Biometrics Ab | Methods of storing a set of biometric data templates and of matching biometrics, biometric matching apparatus and computer program |
| CN104036269B (en) * | 2014-07-03 | 2018-04-17 | 南昌欧菲生物识别技术有限公司 | Fingerprint register method and terminal device |
| CN104331715B (en) * | 2014-10-08 | 2018-08-28 | 清华大学 | Fingerprint posture antidote based on Template Learning and system |
| CN105260696B (en) * | 2015-02-13 | 2020-10-23 | 深圳比亚迪微电子有限公司 | Self-learning method and device of fingerprint template |
| CN105654026A (en) * | 2015-07-16 | 2016-06-08 | 宇龙计算机通信科技(深圳)有限公司 | Fingerprint storage method and apparatus thereof, fingerprint identification method and apparatus thereof |
| CN105808747A (en) * | 2016-03-14 | 2016-07-27 | 浪潮（苏州）金融技术服务有限公司 | Method for quickly searching and comparing fingerprint data by using multidimensional technology |
| US9773147B1 (en) * | 2016-03-25 | 2017-09-26 | Novatek Microelectronics Corp. | Fingerprint enrollment method and apparatus using the same |
| CN108073885B (en) * | 2016-11-18 | 2021-11-12 | 比亚迪半导体股份有限公司 | Fingerprint identification method and electronic device |
| CN107016334A (en) * | 2016-12-23 | 2017-08-04 | 努比亚技术有限公司 | Pattern recognition device and method |
| CN108319883B (en) * | 2017-01-16 | 2020-11-06 | 广东精点数据科技股份有限公司 | Fingerprint identification method based on rapid independent component analysis |
| EP3355239A1 (en) * | 2017-01-27 | 2018-08-01 | Nxp B.V. | Fingerprint verification device |
| CN107392847B (en) * | 2017-06-07 | 2021-01-22 | 西安电子科技大学 | Fingerprint image splicing method based on minutiae and distance images |
| CN108052877B (en) * | 2017-11-28 | 2020-08-07 | Oppo广东移动通信有限公司 | Optical fingerprint identification method and device and electronic equipment |
| CN108171846A (en) * | 2017-12-30 | 2018-06-15 | 南京陶特思软件科技有限公司 | There is the access control system of fast verification |
| CN111222367B (en) * | 2018-11-26 | 2023-11-10 | 上海耕岩智能科技有限公司 | Fingerprint identification method and device, storage medium and terminal |
| CN109891431B (en) * | 2019-01-14 | 2023-05-05 | 深圳市汇顶科技股份有限公司 | Fingerprint identification method, fingerprint identification system and electronic equipment based on multiple security environments |
| CN110199295A (en) * | 2019-04-04 | 2019-09-03 | 深圳市汇顶科技股份有限公司 | The method, apparatus and electronic equipment of fingerprint recognition |
| CN110110714A (en) * | 2019-04-28 | 2019-08-09 | 重庆学析优科技有限公司 | Method and system are corrected automatically on a kind of line of papery operation |
| CN110298305A (en) * | 2019-06-27 | 2019-10-01 | 维沃移动通信有限公司 | A kind of fingerprint identification method and terminal |
| CN112597978B (en) * | 2021-03-03 | 2021-06-22 | 深圳阜时科技有限公司 | Fingerprint matching method and device, electronic equipment and storage medium |
| CN112990163B (en) * | 2021-05-18 | 2021-08-06 | 深圳阜时科技有限公司 | Fingerprint calibration method, electronic device and storage medium |
| CN116311396B (en) * | 2022-08-18 | 2023-12-12 | 荣耀终端有限公司 | Method and device for fingerprint identification |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2310522B (en) * | 1993-04-21 | 1997-10-15 | Matsumura Electronics Kk | Fingerprint ID system and method |
| US20020031245A1 (en) * | 1999-05-14 | 2002-03-14 | Roman Rozenberg | Biometric authentification method |
-
2002
- 2002-12-06 CN CNB021545219A patent/CN100414558C/en not_active Expired - Lifetime
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2310522B (en) * | 1993-04-21 | 1997-10-15 | Matsumura Electronics Kk | Fingerprint ID system and method |
| US20020031245A1 (en) * | 1999-05-14 | 2002-03-14 | Roman Rozenberg | Biometric authentification method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1506903A (en) | 2004-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100414558C (en) | Automatic fingerprint distinguishing system and method based on template learning | |
| You et al. | On hierarchical palmprint coding with multiple features for personal identification in large databases | |
| CN100356388C (en) | Biocharacteristics fusioned identity distinguishing and identification method | |
| EP1467308B1 (en) | Image identification system | |
| Paulino et al. | Latent fingerprint matching using descriptor-based hough transform | |
| Zhang et al. | Selecting a reference high resolution for fingerprint recognition using minutiae and pores | |
| Xu et al. | Fingerprint verification using spectral minutiae representations | |
| Krish et al. | Improving automated latent fingerprint identification using extended minutia types | |
| US20050058325A1 (en) | Fingerprint verification | |
| CN104123537A (en) | Rapid authentication method based on handshape and palmprint recognition | |
| CN100385451C (en) | Deformed fingerprint identification method based on local triangle structure characteristic collection | |
| Sudiro et al. | Adaptable fingerprint minutiae extraction algorithm based-on crossing number method for hardware implementation using FPGA device | |
| Krishneswari et al. | A review on palm print verification system | |
| Bhanu et al. | Human ear recognition by computer | |
| Oldal et al. | Hand geometry and palmprint-based authentication using image processing | |
| TW200813860A (en) | Method and apparatus for adaptive hierarchical processing of print images | |
| Agarwal et al. | A utility of pores as level 3 features in latent fingerprint identification | |
| Daramola et al. | Algorithm for fingerprint verification system | |
| CN117095436A (en) | Intelligent management system and method for enterprise employee information | |
| CN103593660A (en) | Palm print recognition method based on cross gradient encoding of image with stable characteristics | |
| Toh et al. | Combining fingerprint and hand-geometry verification decisions | |
| Liu et al. | A feedback paradigm for latent fingerprint matching | |
| Verma et al. | Static Signature Recognition System for User Authentication Based Two Level Cog, Hough Tranform and Neural Network | |
| US20080240522A1 (en) | Fingerprint Authentication Method Involving Movement of Control Points | |
| Kamaraju et al. | DSP based embedded fingerprint recognition system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C41 | Transfer of patent application or patent right or utility model | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20151111 Address after: 100088 Beijing North Third Ring Road No. 31 Building No. 4 Building 12 layer Taisite Patentee after: CHIPONE TECHNOLOGY(BEIJING) Co.,Ltd. Address before: 100080, No. 1, South Haidian District, Beijing, Zhongguancun Patentee before: Institute of Automation, Chinese Academy of Sciences |
|
| C56 | Change in the name or address of the patentee | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 100088 Beijing North Third Ring Road No. 31 Building No. 4 Building 12 layer Taisite Patentee after: CHIPONE TECHNOLOGY (BEIJING) Co.,Ltd. Address before: 100088 Beijing North Third Ring Road No. 31 Building No. 4 Building 12 layer Taisite Patentee before: CHIPONE TECHNOLOGY(BEIJING) Co.,Ltd. |
|
| CX01 | Expiry of patent term | ||
| CX01 | Expiry of patent term |
Granted publication date: 20080827 |