CN1521684A - Synergic fingerprint identification method - Google Patents
Synergic fingerprint identification method Download PDFInfo
- Publication number
- CN1521684A CN1521684A CNA031128408A CN03112840A CN1521684A CN 1521684 A CN1521684 A CN 1521684A CN A031128408 A CNA031128408 A CN A031128408A CN 03112840 A CN03112840 A CN 03112840A CN 1521684 A CN1521684 A CN 1521684A
- Authority
- CN
- China
- Prior art keywords
- fingerprint
- classification
- vector
- collaborative
- prototype
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Collating Specific Patterns (AREA)
Abstract
A synergic fingerprint identification method characterized in that the overall finger-print image is used as the arithmetic element, a grading group retrieval mode is employed on the basis of synergic finger-print categorization for proceeding synergic finger-print matching, the method by the invention realizes low cost, low complexity, quick identification speed and strong robustness, it has shown high discrimination rate for finger-print sample of noise, microscale rotating, defilement and scraps.
Description
Technical field:
The present invention relates to biometrics identification technology, particularly fingerprint identification technology.
Background technology:
Along with the development of computer and network technology, it is very important that the security of information just seems, biometrics identification technology, and especially the fingerprint recognition technology more and more receives publicity in this field.
In automatic fingerprint recognition, fingerprint classification and coupling are emphasis.Traditional fingerprint classification algorithm mainly is based on the calculating of the field of direction, and traditional fingerprint matching algorithm generally depends on the fingerprint image center, discerns based on the minutia of fingerprint.There are the following problems based on this existing fingerprint identification method:
1, computing machine complexity height, recognition speed are slow.Big multi-method is a main flow with the spatial character of the image slices vegetarian refreshments being calculated one by one with analytical characteristic point, finish search with pixel-wise, whole process need is made repeatedly the traversing operation of entire image, and system is consuming time and calculation cost is quite high, pre-service and feature extraction complexity.
2, for incomplete, stained fingerprint recognition rate is not high, robustness is not strong.In the actual image acquisition process, because the orientation of each finger presses, dynamics are not exclusively the same, can cause fingerprint image existence rotation and distortion in various degree, what other factors caused in addition is fuzzy, incomplete, traditional pre-service and feature extracting method have certain difficulty, and recognition effect is bad.
3, be basic processing unit with single pixel, the Global Information of image is left in the basket, searching image characteristic point from bottom to top, and the more priori that depends in the computation process causes the adaptive ability of classification and coupling not strong.
Summary of the invention:
Technical matters to be solved by this invention is to avoid existing weak point in the above-mentioned prior art, and the collaborative fingerprint identification method that a kind of computation complexity is low, recognition speed is high, the algorithm robustness is stronger is provided.
The technical scheme that technical solution problem of the present invention is adopted is:
The characteristics of the inventive method be with the view picture fingerprint image as arithmetic element, adopt grading group health check-up rope mode, promptly at first work in coordination with fingerprint classification, on the basis of working in coordination with fingerprint classification, work in coordination with fingerprint matching again, concrete steps are:
A, storage: gather the original fingerprint that the access entitlements individuality is arranged, deposit fingerprint database in by arch, left-handed, dextrorotation, cusped arch and whirlpool five big classes;
B, cluster: as learning sample, adopt K-mean cluster method to produce cluster centre, as such other prototype vector with the original fingerprint in each classification;
C, classification: import fingerprint to be identified, be divided in the corresponding classification to be matched by collaborative fingerprint classification method;
D, coupling: as second level prototype vector, carry out fingerprint matching with the original fingerprint stored in this classification, provide recognition result by collaborative fingerprint matching method.
Late 1980s, professor Haken proposes synergy principle is applied to the new ideas of pattern-recognition.The Synergetic Pattern Recognition process can be understood as some preface parameter process of competition, treats recognition mode q and can construct a dynamic process, makes q enter into a pattern v of all prototype patterns through an intermediateness q (t)
k, promptly this prototype pattern and q (0) are the most close.The equation that satisfies pattern-recognition is:
In the formula: λ
kBe attention parameters, have only when it for timing, pattern just can be identified; Q is a pattern to be identified; v
kBe prototype pattern; v
k +Be v
kThe quadrature adjoint vector; F (t) is a fluctuating force.The introducing of preface parameter can make the expression of Synergetic Pattern Recognition process simplify, definition preface parameter
And obedience equation:
In Haken image recognition model, the digital picture matrix at first is converted into the gradation of image value matrix of bidimensional, after the tiling, carries out normalization, zero-mean processing, and transposition is that a dimensional vector is as input vector then.Do not consider spatial information, so state vector is only relevant with time coordinate t.
Collaborative fingerprint classification method can be divided into two processes: the one, ask for the prototype vector that cluster centre obtains five classifications, and the 2nd, the fingerprint assorting process.And collaborative fingerprint matching method also is divided into two processes: the one, and training process, the 2nd, concrete matching process.
Collaborative fingerprint classification and coupling all are to be arithmetic element with the synergetic neural network, and what they were followed is same dynamic process and principle of work, and collaborative fingerprint recognition promptly is to build on this basis.Utilize hierarchical network that automatic fingerprint recognition is achieved.
Above-mentioned model has embodied hierarchical principle, overall situation competition is converted to part competition among several subdomains, each prototype vector number that participates in only is about 1/5th (fingerprint is divided into five major types) of classical Haken network, has overcome the difficulty that adjoint vector was asked for when legacy network was handled the magnanimity pattern-recognition.
In order to overcome the pseudo-situation of recognizing that may occur in WTA in the Haken network (the Winner Take A11) strategy, can adopt preface parameter selection mechanism, prevent the invasion of illegal fingerprint: set a preface parameter introductory die value minimum threshold, if in the fingerprint matching process, maximum introductory die value is lower than this threshold value in the preface parameter, then provide the refusal identifying information, the fingerprint image identification that similarity is too low is unsafe, is illegal.
The collaborative fingerprint characteristic that extracts based on Synergetic Pattern Recognition among the present invention has two globalities: ask for globality and contrast globality.Such compute mode need not traditional characteristic extract in pixel ergodic process one by one, algorithm complex and computing cost decrease.Simultaneously, also preserve the Global Information of image, helped the fingerprint recognition that incompleteness, pixel domain structural information such as stained are lost.It not only extracts the own feature of single image, and considers the difference of this image and other images.Synergizing method does not need shape, texture, the unique point of each image in the database are analyzed separately directly at the general image pixel grey scale, is a kind of indexing means that is independent of outside the application.Synergetic neural network is top-down neural network, and Synergistic method is colony's retrieval mode.Therefore, compared with the prior art, beneficial effect of the present invention is embodied in:
1, the algorithm computation cost of using in the inventive method is little, complexity is low, recognition speed is fast, need not do a large amount of pre-service and complex features extraction work to fingerprint image.
2, the inventive method is stronger to incompleteness, stained sample discrimination height, algorithm robustness, and is not high to the quality requirements of input picture.
3, identification is finished in the inventive method classification, is top-down, does not have the appearance of pseudo-state, and has utilized the Global Information of image, exceeds the dependence priori, and the identification adaptivity is strong, more near the mechanism of the cognitive mechanism of human brain.
4, the inventive method is used for carrying out the occasion that security is taken precautions against as embedded system applicable to network cipher authentication, gate control system etc.Have broad application prospects in fields such as comprising financial instrument, IT industry, security protection, public security, medical treatment, welfare.
The drawing explanation:
Fig. 1 is for implementing the system chart of the inventive method.
Fig. 2 is the collaborative fingerprint classification method block diagram of the present invention.
Fig. 3 is the collaborative fingerprint matching method block diagram of the present invention.
Embodiment:
As arithmetic element, adopt grading group health check-up rope mode with the view picture fingerprint image, promptly at first adopt collaborative fingerprint classification device, by five big classes of fingerprint: encircle, left-handed, dextrorotation, cusped arch and whirlpool five big classes classify, promptly coarse grade is discerned fast; Then, on the result of collaborative fingerprint classification, in a certain collaborative fingerprint matching device, mate the identification of promptly meticulous level.Concrete steps are:
A, storage: gather the original fingerprint that the access entitlements individuality is arranged, deposit fingerprint database in by arch, left-handed, dextrorotation, cusped arch and whirlpool five big classes;
B, cluster: as learning sample, adopt K-mean cluster method to produce cluster centre, as first order classification prototype vector with the original fingerprint in each classification;
C, classification: import fingerprint to be identified, divide in the respective classes it to be matched by collaborative fingerprint classification method;
D, coupling: as second level prototype vector, carry out fingerprint matching with the original fingerprint stored in this classification, provide recognition result by collaborative fingerprint matching method.
Referring to Fig. 1, the system that implements the inventive method mainly is made up of two parts: collaborative fingerprint classification device and collaborative fingerprint matching device.Wherein, collaborative fingerprint classification device is made up of pre-stored system and categorizing system, and collaborative fingerprint classification method block diagram as shown in Figure 2.Collaborative fingerprint matching device is made up of pre-stored system and matching system, and collaborative fingerprint matching method block diagram as shown in Figure 3.
Referring to Fig. 2, the specific implementation of collaborative fingerprint classification method is:
Choose a certain amount of representative sample in a, each classification, adopt the K-means clustering algorithm to obtain this classification cluster centre;
B, with after the five classification cluster centre image vectorizations, handling, as prototype vector v through normalization and zero-mean
kDeposit the pre-stored matrix in;
C, obtain prototype vector v
kAdjoint vector v
k +, finish e-learning;
D, fingerprint image vectorization to be identified calculate the input mode vector q (0) that satisfies normalization and zero-mean condition;
E, by input mode vector q (0) and adjoint vector v
k +, obtain the initial value ξ of preface parameter
k(0);
F, each preface parameter competition are developed, and in certain prototype pattern, fingerprint to be identified is included into the classification of this prototype pattern representative up to system stability.In collaborative fingerprint classification, employed preface parameter competition equation is:
Wherein γ is an iteration step length.
Referring to Fig. 3, work in coordination with being embodied as of fingerprint matching method:
A, training image vectorization calculate the prototype pattern vector v that satisfies normalization and zero-mean
k
B, obtain prototype vector v
kAdjoint vector v
k +
C, image vectorization to be identified calculate the input mode vector q (0) that satisfies normalization and zero-mean condition;
D, by input mode vector q (0) and adjoint vector v
k +, obtain the initial value of preface parameter; Check input sample legitimacy: calculate the initial value of each pattern preface parameter, judge whether the input fingerprint is legal:, then change step e if legal; Otherwise, then enter step f;
E, the preface parameter is developed, identify fingerprint pattern in certain pattern up to system stability;
F, provide illegal input information, refusal identification finishes.
Applicating example:
Suppose certain LAN system, individual N of access entitlements arranged, and each individuality can only enter corresponding separately subsystem.As the authentication means, adopt collaborative fingerprint recognition system with fingerprint recognition, can followingly realize.
At first, system gathers N individual training sample fingerprint a respectively
1, a
2..., a
NIndividual, be total to M training sample.This M training sample is divided into 5 classes, and the cluster centre vector of each classification is as the prototype vector of collaborative fingerprint classification device, and the original fingerprint sample of each classification, after the vectorization, as the prototype vector of corresponding fingerprint matching device.After system finishes training, just can discern.
During system identification, sample to certain individuality to be identified of gathering, at first in the fingerprint classification device, classify, by collaborative fingerprint classification algorithm it is divided into a certain classification, enter corresponding fingerprint matching device, carry out fingerprint matching according to collaborative fingerprint matching algorithm again, provide recognition result, individuality can enter corresponding subsystem.If individuality to be identified is not that some in N the individuality of access entitlements arranged, system will provide illegal input information, and refusal is discerned.
Claims (3)
1, collaborative fingerprint identification method is characterized in that with the view picture fingerprint image adopting grading group health check-up rope mode as arithmetic element, promptly at first work in coordination with fingerprint classification, on the basis of working in coordination with fingerprint classification, works in coordination with fingerprint matching again, and concrete steps are:
(a), storage: gather the original fingerprint that the access entitlements individuality is arranged, deposit fingerprint database in by arch, left-handed, dextrorotation, cusped arch and whirlpool five big classes;
(b), cluster: as learning sample, adopt K-mean cluster method to produce cluster centre, as first order classification prototype vector with the original fingerprint in each classification;
(c), classification: import fingerprint to be identified, divide in the respective classes it to be matched by collaborative fingerprint classification method;
(d), coupling: as second level prototype vector, carry out fingerprint matching with the original fingerprint stored in this classification, provide recognition result by collaborative fingerprint matching method.
2, collaborative fingerprint identification method according to claim 1 is characterized in that described collaborative fingerprint classification method is:
(a), choose a certain amount of representative sample in each classification, adopt the K-means clustering algorithm to obtain this classification cluster centre;
(b), with after the five classification cluster centre image vectorizations, handling, as prototype vector v through normalization and zero-mean
kDeposit the pre-stored matrix in;
(c), obtain prototype vector v
kAdjoint vector v
k +, finish e-learning;
(d), fingerprint image vectorization to be identified, calculate the input mode vector q (0) that satisfies normalization and zero-mean condition;
(e), by input mode vector q (0) and adjoint vector v
k +, obtain the initial value ξ of preface parameter
k(0);
(f), the competition of each preface parameter develops, in certain prototype pattern, fingerprint to be identified is included into the classification of this prototype pattern representative up to system stability.
3, collaborative fingerprint identification method according to claim 1 is characterized in that described collaborative fingerprint matching method is:
(a), the training image vectorization, calculate the prototype pattern vector v that satisfies normalization and zero-mean
k
(b), obtain prototype vector v
kAdjoint vector v
k +
(c), image vectorization to be identified, calculate the input mode vector q (0) that satisfies normalization and zero-mean condition;
(d), by input mode vector q (0) and adjoint vector v
k +, obtain the initial value of preface parameter; Check input sample legitimacy: calculate the initial value of each pattern preface parameter, judge whether the input fingerprint is legal:, then change step e if legal; Otherwise, then enter step f;
(e), the preface parameter is developed, identify fingerprint pattern up to system stability in certain pattern;
(f), provide illegal input information, refusal identification, end.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB031128408A CN1246800C (en) | 2003-02-13 | 2003-02-13 | Synergic fingerprint identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB031128408A CN1246800C (en) | 2003-02-13 | 2003-02-13 | Synergic fingerprint identification method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1521684A true CN1521684A (en) | 2004-08-18 |
CN1246800C CN1246800C (en) | 2006-03-22 |
Family
ID=34283660
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB031128408A Expired - Fee Related CN1246800C (en) | 2003-02-13 | 2003-02-13 | Synergic fingerprint identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN1246800C (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1322464C (en) * | 2003-09-24 | 2007-06-20 | 三洋电机株式会社 | Authentication apparatus and authentication method |
CN100460813C (en) * | 2007-05-10 | 2009-02-11 | 上海交通大学 | Three-D connection rod curve matching rate detection method |
WO2010069166A1 (en) * | 2008-12-19 | 2010-06-24 | 杭州中正生物认证技术有限公司 | Fast fingerprint searching method and fast fingerprint searching system |
CN102273128A (en) * | 2008-12-08 | 2011-12-07 | 茂福公司 | Identification or authorisation method, and associated system and secure module |
CN105608409A (en) * | 2015-07-16 | 2016-05-25 | 宇龙计算机通信科技(深圳)有限公司 | Method and device for fingerprint identification |
CN106485125A (en) * | 2016-10-21 | 2017-03-08 | 上海与德信息技术有限公司 | A kind of fingerprint identification method and device |
CN106557761A (en) * | 2016-11-29 | 2017-04-05 | 深圳天珑无线科技有限公司 | A kind of method and device of fingerprint recognition |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874899B (en) * | 2017-04-20 | 2018-10-19 | 维沃移动通信有限公司 | A kind of fingerprint identification method and mobile terminal |
-
2003
- 2003-02-13 CN CNB031128408A patent/CN1246800C/en not_active Expired - Fee Related
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1322464C (en) * | 2003-09-24 | 2007-06-20 | 三洋电机株式会社 | Authentication apparatus and authentication method |
CN100460813C (en) * | 2007-05-10 | 2009-02-11 | 上海交通大学 | Three-D connection rod curve matching rate detection method |
CN102273128A (en) * | 2008-12-08 | 2011-12-07 | 茂福公司 | Identification or authorisation method, and associated system and secure module |
CN102273128B (en) * | 2008-12-08 | 2015-07-15 | 茂福公司 | Identification or authorisation method, and associated system and secure module |
WO2010069166A1 (en) * | 2008-12-19 | 2010-06-24 | 杭州中正生物认证技术有限公司 | Fast fingerprint searching method and fast fingerprint searching system |
CN105608409A (en) * | 2015-07-16 | 2016-05-25 | 宇龙计算机通信科技(深圳)有限公司 | Method and device for fingerprint identification |
WO2017008349A1 (en) * | 2015-07-16 | 2017-01-19 | 宇龙计算机通信科技(深圳)有限公司 | Fingerprint recognition method and device |
CN105608409B (en) * | 2015-07-16 | 2019-01-11 | 宇龙计算机通信科技(深圳)有限公司 | The method and device of fingerprint recognition |
CN106485125A (en) * | 2016-10-21 | 2017-03-08 | 上海与德信息技术有限公司 | A kind of fingerprint identification method and device |
CN106485125B (en) * | 2016-10-21 | 2020-05-05 | 深圳市尚优像电子有限公司 | Fingerprint identification method and device |
CN106557761A (en) * | 2016-11-29 | 2017-04-05 | 深圳天珑无线科技有限公司 | A kind of method and device of fingerprint recognition |
Also Published As
Publication number | Publication date |
---|---|
CN1246800C (en) | 2006-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Cui et al. | Identifying materials of photographic images and photorealistic computer generated graphics based on deep CNNs. | |
Doegar et al. | Cnn based image forgery detection using pre-trained alexnet model | |
CN104537393B (en) | A kind of traffic sign recognition method based on multiresolution convolutional neural networks | |
Kashyap et al. | Automatic number plate recognition | |
Agarwal et al. | Hyperfeatures–multilevel local coding for visual recognition | |
Dogantekin et al. | A robust technique based on invariant moments–ANFIS for recognition of human parasite eggs in microscopic images | |
Arsa et al. | VGG16 in batik classification based on random forest | |
CN1908960A (en) | Feature classification based multiple classifiers combined people face recognition method | |
CN1658219A (en) | Image processing apparatus | |
CN109740603A (en) | Based on the vehicle character identifying method under CNN convolutional neural networks | |
CN101034433A (en) | Method for classification human facial expression and semantics judgement quantization method | |
Ahmed et al. | Directional ternary pattern (DTP) for facial expression recognition | |
CN105160305B (en) | A kind of multi-modal Feature fusion of finger | |
Kancharla et al. | Handwritten signature recognition: a convolutional neural network approach | |
CN1246800C (en) | Synergic fingerprint identification method | |
CN112580445A (en) | Human body gait image visual angle conversion method based on generation of confrontation network | |
Agarwal et al. | Multilevel image coding with hyperfeatures | |
Althabhawee et al. | Fingerprint recognition based on collected images using deep learning technology | |
Zhou et al. | Fast minutiae extractor using neural network | |
CN117392375A (en) | Target detection algorithm for tiny objects | |
CN1858773A (en) | Image identifying method based on Gabor phase mode | |
Ganapathi et al. | A fuzzy hybrid framework for offline signature verification | |
CN1144158C (en) | Method for identifying human body biological characteristics | |
CN101826208A (en) | Image segmentation method combining support vector machine and region growing | |
Amin et al. | Hand-printed character recognition system using artificial neural networks |
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 | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20060322 Termination date: 20100213 |