CN103839072B - False fingerprint detecting method based on naive Bayes classifiers - Google Patents
False fingerprint detecting method based on naive Bayes classifiers Download PDFInfo
- Publication number
- CN103839072B CN103839072B CN201310754757.5A CN201310754757A CN103839072B CN 103839072 B CN103839072 B CN 103839072B CN 201310754757 A CN201310754757 A CN 201310754757A CN 103839072 B CN103839072 B CN 103839072B
- Authority
- CN
- China
- Prior art keywords
- fingerprint
- image
- feature
- false
- graders
- 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.)
- Active
Links
Landscapes
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
Abstract
A false fingerprint detecting method based on naive Bayes classifiers comprises the following steps of (1) training library dividing; (2) image normalization; (3) feature extraction, wherein the feature extraction comprises (3.1) discrete wavelet transformation, (3.2) noise removal, (3.3) wavelet reconstruction, (3.4) noise pattern estimation, (3.5) standard deviation chart calculation, and (3.6) standard deviation chart dividing, and image feature obtaining through statistic conducting; (4) feature dividing; (5) classifier training; (6) classifier performance estimation; (7) classifier fusion: utilizing a construction method of the naive Bayes classifiers for conducting fusion to obtain a new classifier. According to the method, the requirement for the performance of a single classifier is not high, but the effect of the classifiers after fusion is quite good.
Description
Technical field
The present invention relates to the technical field such as image procossing, machine learning, pattern recognition, especially one kind are directed to fingerprint image
True and false detection method.
Background technology
Image procossing, feature extraction, classifier training and image classification etc. are the important steps in false fingerprint detection method.
Y.S.Moon etc. proposes a kind of false fingerprint detection algorithm based on noise, and is 1000dpi to 200 resolution
True and false fingerprint image tested, all classification is correct.L.F.A.Pereria etc. according to the paper of Moon, it is further proposed that
The false fingerprint detection algorithm based on noise of the fingerprint image being 500dpi for resolution, and to LivDet2011's
Sagem is tested in storehouse, and averagely sentencing error rate is 12.8%.
The principle of the detection method based on noise, is described in the paper of Y.S.Moon, and general principle is:The system of main flow
Play tricks the material such as clay and gelatin of fingerprint, during making false fingerprint, due to comprising big organic molecule, typically can condense
In bulk, which results in the coarse of false fingerprint.Therefore, in general, the finger tip surface of false fingerprint is more more coarse than live body.This
One characteristic can be used as the foundation of false fingerprint detection algorithm, its showing as in data:The noise ratio of false fingerprint image really refers to
The noise of print image has bigger standard deviation.Algorithm due to proposing in Moon paper is directed to the fingerprint that resolution is 1000dpi
Image, and commonly use at present be 500dpi fingerprint collecting device, in order to check the suitability to 500dpi image for its algorithm,
L.F.A.Pereria is tested with the sagem fingerprint base of LivDet2011, and averagely sentencing error rate is 42.8%, with current main flow
Algorithm performance differs greatly.Therefore L.F.A.Pereria etc. this algorithm has been carried out improve it is proposed that based on spatial surface coarse
Degree analysis(Spatial surface coarseness analysis, abbreviation SSCA)False fingerprint detection method, mainly to spy
Levy extraction to be improved, by noise pattern piecemeal and seek standard deviation, to the standard deviation figure piecemeal obtaining and count, last statistics
Value its feature the most, comprehensive all fritters, that is, obtain the feature of fingerprint image.This algorithm is in the sagem fingerprint of LivDet2011
What in image library, test obtained averagely sentences error rate is 12.8%.
Support vector machine are a kind of methods of supervised study, can be widely used in statistical classification and regression analyses.
Support vector machine belong to vague generalization linear classifier, are also considered as proposing clo husband standardization(Tikhonov
Regularization)One special case of method.The feature of this grader is can to minimize experience error and maximum simultaneously
Change Geometry edge area, therefore support vector machine are also referred to as maximal margin region class device.
Naive Bayes Classifier is a kind of Bayesian simple probability grader based on independent hypothesis for application,
It is independent characteristic model that this potential probabilistic model can more accurately be described.The basis of Bayes's classification is probability inference, just
It is in the probability of occurrence only knowing each condition, and in the case that uncertain its whether there is, how to complete reasoning and decision task.And
Naive Bayes Classifier is to be assumed based on independent, that is, assume that each feature of sample is uncorrelated to other features, even if this
A little features interdepend or some features are determined by other features, all think that these attributes are independent in decision probability distribution
's.Naive Bayes Classifier relies on accurate natural probability model, and it is very good to obtain in the sample of supervised learning
Classifying quality.In many practical applications, model-naive Bayesian parameter estimation uses maximum Likelihood, not
Use Bayesian probability or any Bayesian model.Despite with simple thought and excessively simplification it is assumed that simple pattra leaves
This grader still is able to obtain goodish effect in much complicated real-world situation.2004, an analysis Bayes divided
The reason article of class device problem theoretically discloses some Naive Bayes Classifiers acquirement mysterious classifying qualities.
One advantage of Naive Bayes Classifier is to only need to just to estimate necessary parameter according to a small amount of training data.Logical
Cross simple probabilistic method to analyze it is recognised that the shortcoming of current false fingerprint detection method generally existing is only to use single classifier
To be classified, but the result of the false fingerprint detection algorithm of single classifier is less reliable.
Content of the invention
In order to make up the insecure deficiency of single grader classification results of existing vacation fingerprint detection method, the present invention provides
A kind of good false fingerprint detection method based on Naive Bayes Classifier of classification results reliability.
The technical solution adopted for the present invention to solve the technical problems is:
A kind of false fingerprint detection method based on Naive Bayes Classifier, comprises the following steps:
1) training storehouse divides:Training storehouse image is randomly divided into equal two part, is designated as A, B;
2) image normalization:Using fingerprint image normalized function, picture size is unified into m pixel × n-pixel(M, n's
Value is 4 multiple, generally can take 256,512 etc.);
3) feature extraction:Noise extraction and process are carried out to image, including procedure below:
3.1)Wavelet transform is carried out to image, obtains a low frequency part and six HFSs;
3.2)To the six HFSs hyperbolic shrinkage method denoising after conversion;
3.3)According to step 3.2)Six HFSs obtaining and low frequency part before carry out wavelet reconstruction, are gone
Image after making an uproar;
3.4)By image subtraction after original image and denoising, obtain noise pattern;
3.5)Noise pattern is divided intoIndividual fritter, wherein pxWidth for fritter and the ratio of original image width, py
Length for fritter and the ratio of original image length, calculate each piece of standard deviation, obtain standard deviation figure;
3.6)Calculate the maximum of standard deviation in figure, be designated as S, interval [0, S] is divided into(It is designated as k)Equal portions,
Obtain k interval, standard deviation figure is divided intoIndividual fritter, wherein qxWidth for fritter and the ratio of original image width,
qyLength for fritter and the ratio of original image length, count the number that each fritter Plays difference falls in k interval, obtain
It is worth to k, in this, as this little block feature, comprehensive each piece, that is, obtain this characteristics of image;
4) feature divides:Feature is divided into 4 parts, corresponds respectively to the upper left of image, upper right, lower-left, bottom right four portion
Point;
5) classifier training:It is respectively trained the feature after division with SVM, obtains grader, for two training storehouses of A, B altogether
Available 8 graders;
6) classifier performance assessment:
6.1)For 4 graders in A, the feature being obtained with correspondence position in B respectively, to be tested, counts vacation and refers to
Sentencing of stricture of vagina sentences error rate to rate and true fingerprint;
6.2)In the same manner, for 4 graders in B, the feature being obtained with correspondence position in A respectively, to be tested, is united
Sentencing of the false fingerprint of meter sentences error rate to rate and true fingerprint;
7) Multiple Classifier Fusion:
7.1)For a unknown fingerprint image it is believed that the probability that it is false fingerprint is prior probability, comprehensive 8 classification
After the classification results of device, when the probability being judged to false fingerprint is more than or equal to probability threshold value T, that is, think that this fingerprint is false fingerprint;
7.2)For 8 graders obtaining before, it is designated as f respectively1、f2, f3, f4, f5, f6, f7, f8, note C is classification collection
Close { 0,1 }, wherein 0 represents false fingerprint, and 1 represents true fingerprint;
7.3)For this 8 graders it is believed that being separate between them, therefore, in the knot of known 8 graders
Really, and 8 graders performance, then use Bayesian formula, note P1=∏ P (C=0 | fi(X)), P2=∏P(C=1|fi(X)),
Obtaining a fingerprint image is false fingerprint probability P=P1/ (P1+P2), when P is more than T, is judged to 0, otherwise for 1, P therein (C=0 |
fi(X)) all can be obtained by Bayesian formula, the grader F after therefore being merged;
8) false fingerprint detection:1 is carried out to image to be detected)、2)、3)The operation of step, then by the characteristic vector obtaining
Classified with grader F.
The technology design of the present invention is:Find when to existing vacation fingerprint detection algorithm paractical research, with single point
Class device image is classified, and the classification results obtaining are less reliable.This problem can be treated from the perspective of probability, real
When detection when it is assumed that fingerprint to be measured be false fingerprint probability be 0.01, have a grader, it is sentenced to general to false fingerprint
Rate is a, and the wrong probability of sentencing to true fingerprint is b, then obtained according to Bayesian formula, test fingerprint is false fingerprint, and is classified
Device is judged to be the probability of false fingerprintIt is further assumed that working as p>It is judged to false fingerprint during prior probability 0.5,
It is possible to obtain a relational expression with regard to a, b:A > 99 × b, because a<1, so b<1/99.For this precision, mesh
The detection method of front single classifier can't reach, the method that therefore just take into account fusion here.And this process can also
To analyze from another angle, single classifier is judged to false fingerprint so that judging that this fingerprint is that the probability of false fingerprint improves, but
This probability is also not enough to judge that it is false fingerprint.According to this thinking it is proposed that fusion method, first determine a series of mutual
Independent grader, using them as the required independent feature of Naive Bayes Classifier construction, with each grader come right
Fingerprint is classified, the result classified for each, can increase or reduce it be false fingerprint probability.Therefore, as long as
There is the sufficient amount of grader with certain classification capacity, with regard to meeting the requirement that true and false fingerprint is classified, this is also
The thought of Naive Bayes Classifier construction.
The present invention, on the basis of the method based on SSCA that L.F.A.Pereria proposes, first constructs multiple independent dividing
Class device, using each grader as false fingerprint image independent characteristic, then with construct Naive Bayes Classifier method merge
Obtain last grader.
Beneficial effects of the present invention are:Less demanding to single classifier performance, but the effect after Multiple Classifier Fusion can
With very good, classification results reliability is good.
Brief description
Fig. 1 is a kind of false fingerprint detection method flow chart based on Naive Bayes Classifier.
Fig. 2 is the flow chart of feature extraction.
Fig. 3 is the schematic diagram of feature extraction.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
With reference to Fig. 1, a kind of false fingerprint detection method based on Naive Bayes Classifier, described fingerprint detection method includes
Following steps:
1) training storehouse divides:Training storehouse image is randomly divided into equal two part, is designated as A, B;
2) image normalization:Using fingerprint image normalized function, picture size is unified into m pixel × n-pixel;
Just there are the concrete steps of following feature extraction with reference to Fig. 2 and Fig. 3.
3) feature extraction:It is process picture noise extracted and processes, including procedure below:
A) two-dimensional discrete wavelet conversion is carried out to image f (x, y), obtain a low frequency part and six HFS gk
(x,y),k∈{1,2,3,4,5,6};
B) to the six HFSs hyperbolic shrinkage method denoising after conversion:
Wherein N is gkThe signal length of (x, y), σ is the standard deviation of 3 details of the first dimension decomposed by conversion;
C) according to step a)Six HFSs obtaining and low frequency part before carry out wavelet reconstruction, after obtaining denoising
Image f'(x, y);
D) by image subtraction after original image and denoising, obtain noise pattern;
E) noise pattern is divided intoIndividual fritter, calculates each piece of standard deviation, obtains standard deviation figure;
F) calculate the maximum of standard deviation in figure, be designated as S, interval [0, S] is divided into(It is designated as k)Equal portions, obtain
Interval to k.Standard deviation figure is divided intoIndividual fritter, counts each fritter Plays difference and falls in k interval
Number, obtains k value, in this, as this little block feature, comprehensive each piece, that is, obtains this characteristics of image;
4) feature divides:Feature is divided into 4 parts, corresponds respectively to the upper left of image, upper right, lower-left, bottom right four portion
Point;
5) classifier training:For the feature of each class, all it is trained with SVM, obtain grader, thus altogether
Obtain 8 graders;
6) classifier performance is estimated:
A) for 4 graders in A, the feature being obtained with correspondence position in B respectively, to be tested, counts false fingerprint
Sentence error rate sentenced to rate and true fingerprint;
B) in the same manner, for 4 graders in B, the feature being obtained with correspondence position in A respectively, to be tested, counts
Sentencing of false fingerprint sentences error rate to rate and true fingerprint;
7) Multiple Classifier Fusion:
A) for a unknown fingerprint image it is considered to it is the probability of false fingerprint(It is prior probability)For 0.5, and
The classification results of comprehensive 8 graders, when the probability being false fingerprint is more than or equal to probability threshold value T it is believed that being false fingerprint;
B) for 8 graders obtaining before, it is designated as f1, f2, f3, f4, f5, f6, f7, f8 respectively, note C is classification collection
Close { 0,1 }, wherein 0 represents false fingerprint, and 1 represents true fingerprint;
C) for this 8 graders it is believed that being separate between them, therefore, in the knot of known 8 graders
Really, and 8 graders performance, then use Bayesian formula, note P1=∏ P (C=0 | fi(X)), P2=∏ P (C=1 | fi(X)),
It is false fingerprint probability that a fingerprint image can be obtained(It is posterior probability)P=P1/ (P1+P2), when P is more than T, is judged to 0, no
It is then 1, and P therein (C=0 | fi(X)) all can be obtained by Bayesian formula, the grader F after therefore being merged;
8) false fingerprint detection:1 is carried out to image to be detected)、2)、3)The operation of step, then by the characteristic vector obtaining
Classified with grader F.
Last table 1 is the test result with this detection method to LivDet2011 match image library.
Table 1.
Claims (1)
1. a kind of false fingerprint detection method based on Naive Bayes Classifier it is characterised in that:Described vacation fingerprint detection method
Comprise the following steps:
1) training storehouse divides:Training storehouse image is randomly divided into equal two part, is designated as A, B;
2) image normalization:Using fingerprint image normalized function, picture size is unified into m pixel × n-pixel, the value of m, n
Multiple for 4;
3) feature extraction:Noise extraction and process are carried out to image, including procedure below:
3.1) wavelet transform is carried out to image, obtain a low frequency part and six HFSs;
3.2) to the six HFSs hyperbolic shrinkage method denoising after conversion;
3.3) according to step 3.2) six HFSs obtaining and low frequency part before carry out wavelet reconstruction, after obtaining denoising
Image;
3.4) by image subtraction after original image and denoising, obtain noise pattern;
3.5) noise pattern is divided intoIndividual fritter, wherein pxWidth for fritter and the ratio of original image width, pyFor little
The length of block and the ratio of original image length, calculate each piece of standard deviation, obtain standard deviation figure;
3.6) calculate the maximum of standard deviation in figure, be designated as S, interval [0, S] is divided into k part,Obtain k area
Between, standard deviation figure is divided intoIndividual fritter, wherein qxWidth for fritter and the ratio of original image width, qyFor fritter
Length and original image length ratio, count each fritter Plays difference fall into k interval in number, obtain k value,
In this, as this little block feature, comprehensive each piece, that is, obtain this characteristics of image;
4) feature divides:Feature is divided into 4 parts, corresponds respectively to the upper left of image, upper right, lower-left, bottom right four part;
5) classifier training:It is respectively trained the feature after division with SVM, obtains grader, can obtain altogether for two training storehouses of A, B
To 8 graders;
6) classifier performance assessment:
6.1) for 4 graders in A, the feature being obtained with correspondence position in B respectively, to be tested, counts false fingerprint
Sentence and error rate is sentenced to rate and true fingerprint;
6.2) in the same manner, for 4 graders in B, the feature being obtained with correspondence position in A respectively, to be tested, counts false
Sentencing of fingerprint sentences error rate to rate and true fingerprint;
7) Multiple Classifier Fusion:
7.1) for a unknown fingerprint image it is believed that the probability that it is false fingerprint is prior probability, comprehensive 8 graders
After classification results, when the probability being judged to false fingerprint is more than or equal to probability threshold value T, that is, think that this fingerprint is false fingerprint;
7.2) for 8 graders obtaining before, it is designated as f respectively1、f2, f3, f4, f5, f6, f7, f8, note C be category set 0,
1 }, wherein 0 represent false fingerprint, 1 represents true fingerprint;
7.3) for this 8 graders it is believed that being separate between them, therefore, in the result of known 8 graders,
And the performance of 8 graders, then use Bayesian formula, and note P1=∏ P (C=0 | fi(X)), P2=∏ P (C=1 | fi
(X)), obtain probability P=P1/ (P1+P2) that fingerprint image is false fingerprint, when P is more than T, be judged to 0, otherwise for 1, wherein
P (C=0 | fi(X)) all can be obtained by Bayesian formula, the grader F after therefore being merged;
8) false fingerprint detection:1 is carried out to image to be detected), 2), 3) operation of step, then by the characteristic vector obtaining with point
Class device F is being classified.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310754757.5A CN103839072B (en) | 2013-12-31 | 2013-12-31 | False fingerprint detecting method based on naive Bayes classifiers |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310754757.5A CN103839072B (en) | 2013-12-31 | 2013-12-31 | False fingerprint detecting method based on naive Bayes classifiers |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103839072A CN103839072A (en) | 2014-06-04 |
CN103839072B true CN103839072B (en) | 2017-02-22 |
Family
ID=50802550
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310754757.5A Active CN103839072B (en) | 2013-12-31 | 2013-12-31 | False fingerprint detecting method based on naive Bayes classifiers |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103839072B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361319B (en) * | 2014-11-10 | 2018-01-09 | 杭州景联文科技有限公司 | A kind of false fingerprint detection method based on SVM RFE feature selectings |
CN104794477B (en) * | 2015-04-27 | 2016-04-13 | 山东大学 | Based on the high spectrum image Feature Extraction Method of 3-D wavelet transformation and sparse tensor |
CN105023025B (en) * | 2015-08-03 | 2018-09-25 | 大连海事大学 | A kind of opener mark image sorting technique and system |
CN106372656B (en) * | 2016-08-30 | 2019-05-10 | 同观科技(深圳)有限公司 | Obtain method, image-recognizing method and the device of the disposable learning model of depth |
CN112241657A (en) | 2019-07-17 | 2021-01-19 | 华为技术有限公司 | Fingerprint anti-counterfeiting method and electronic equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101226589A (en) * | 2007-01-18 | 2008-07-23 | 中国科学院自动化研究所 | Method for detecting living body fingerprint based on thin plate spline deformation model |
CN103116744A (en) * | 2013-02-05 | 2013-05-22 | 浙江工业大学 | Fake fingerprint detection method based on markov random field (MRF) and support vector machine-k nearest neighbor (SVM-KNN) classification |
CN103324944A (en) * | 2013-06-26 | 2013-09-25 | 电子科技大学 | Fake fingerprint detecting method based on SVM and sparse representation |
-
2013
- 2013-12-31 CN CN201310754757.5A patent/CN103839072B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101226589A (en) * | 2007-01-18 | 2008-07-23 | 中国科学院自动化研究所 | Method for detecting living body fingerprint based on thin plate spline deformation model |
CN103116744A (en) * | 2013-02-05 | 2013-05-22 | 浙江工业大学 | Fake fingerprint detection method based on markov random field (MRF) and support vector machine-k nearest neighbor (SVM-KNN) classification |
CN103324944A (en) * | 2013-06-26 | 2013-09-25 | 电子科技大学 | Fake fingerprint detecting method based on SVM and sparse representation |
Also Published As
Publication number | Publication date |
---|---|
CN103839072A (en) | 2014-06-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Al-masni et al. | Detection and classification of the breast abnormalities in digital mammograms via regional convolutional neural network | |
Juszczak et al. | Minimum spanning tree based one-class classifier | |
CN103839072B (en) | False fingerprint detecting method based on naive Bayes classifiers | |
CN105844279B (en) | SAR image change detection based on deep learning and SIFT feature | |
CN106203521B (en) | The SAR image change detection learnt based on disparity map from step | |
CN103699874B (en) | Crowd abnormal behavior identification method based on SURF (Speed-Up Robust Feature) stream and LLE (Locally Linear Embedding) sparse representation | |
CN111291817B (en) | Image recognition method, image recognition device, electronic equipment and computer readable medium | |
CN110263712A (en) | A kind of coarse-fine pedestrian detection method based on region candidate | |
CN111009321A (en) | Application method of machine learning classification model in juvenile autism auxiliary diagnosis | |
CN111950488B (en) | Improved Faster-RCNN remote sensing image target detection method | |
CN108171119B (en) | SAR image change detection method based on residual error network | |
CN104765846A (en) | Data feature classifying method based on feature extraction algorithm | |
CN109817339A (en) | Patient's group technology and device based on big data | |
Vu et al. | Energy-based models for video anomaly detection | |
Mohammed et al. | Breast tumor classification using a new OWA operator | |
CN107609588A (en) | A kind of disturbances in patients with Parkinson disease UPDRS score Forecasting Methodologies based on voice signal | |
Luo et al. | Traffic analytics with low-frame-rate videos | |
Li et al. | Robust outlier detection based on the changing rate of directed density ratio | |
Henila et al. | Segmentation using fuzzy cluster‐based thresholding method for apple fruit sorting | |
Djunaidi et al. | Gray level co-occurrence matrix feature extraction and histogram in breast cancer classification with ultrasonographic imagery | |
Asodekar et al. | Brain tumor classification using shape analysis of MRI images | |
Fernández-Carrobles et al. | Bagging tree classifier and texture features for tumor identification in histological images | |
Lust et al. | A survey on assessing the generalization envelope of deep neural networks: predictive uncertainty, out-of-distribution and adversarial samples | |
Sannasi Chakravarthy et al. | Performance analysis of ensemble classifiers and a two-level classifier in the classification of severity in digital mammograms | |
Nancy et al. | Optimized Feature Selection and Image Processing Based Machine Learning Technique for Lung Cancer Detection |
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 |