CN108921186A - A kind of fingerprint image categorizing system and method based on twin support vector machines - Google Patents
A kind of fingerprint image categorizing system and method based on twin support vector machines Download PDFInfo
- Publication number
- CN108921186A CN108921186A CN201810429965.0A CN201810429965A CN108921186A CN 108921186 A CN108921186 A CN 108921186A CN 201810429965 A CN201810429965 A CN 201810429965A CN 108921186 A CN108921186 A CN 108921186A
- Authority
- CN
- China
- Prior art keywords
- fingerprint image
- support vector
- twin
- fingerprint
- vector machines
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1347—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1365—Matching; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses a kind of fingerprint image categorizing systems and method based on twin support vector machines, it is related to quanta particle swarm optimization and " one-to-many " classifies twin algorithm of support vector machine more, the parameter that can more classify to " one-to-many " in twin support vector machine classifier optimizes, the classification for completing fingerprint image, belongs to artificial intelligence and image classification field.The invention mainly comprises following contents:Step 1:Fingerprint image preprocessing;Step 2:The feature extraction of fingerprint image;Step 3:" one-to-many " more classify parameter of twin support vector machine classifier is optimized with quanta particle swarm optimization;Step 4:It brings optimized parameter into " one-to-many " to classify in twin support vector machine classifier more, determines final disaggregated model and classify to fingerprint image.Twin support vector machines is successfully applied to fingerprint classification and obtains higher classification accuracy by this method.
Description
Technical field
A kind of fingerprint image categorizing system and method based on twin support vector machines of the present invention is related to quantum particle swarm calculation
Method and " one-to-many " are classified twin algorithm of support vector machine more, and can classify to " one-to-many " twin support vector machine classifier more
In parameter optimize, complete the classification of fingerprint image, belong to artificial intelligence and image classification field.
Background technique
Currently, fingerprint identification technology has been widely used in the fields such as attendance, gate inhibition, e-commerce, mobile phone safe.By
In the importance of fingerprint identification technology, which has become one, machine learning field important research content.Fingerprint image point
Class is one of core procedure of fingerprint recognition, is directly related to the accuracy rate of identification, to realize magnanimity fingerprint image retrieval and
Establish perfect Indexing Mechanism all play the role of it is vital.
Under normal conditions, fingerprint can be divided into six classes:Left dustpan type (Left Loop), right dustpan type (Right Loop), bucket type
(Whorl), arcuate (Arch), sharp arcuate (TentedArch), double dustpan types (Twin Loops).
In recent decades, the fingerprint image sorting algorithm of many classics is developed both at home and abroad.These methods can substantially divide
For four major class, i.e.,:Method, structure-based method based on model, the method based on frequency and the method based on grammer.Its
In, the algorithm of structure-based Fingerprint diretion is method most popular in fingerprint classification.Such as:Cappelli et al. [1]
The fingerprint classification algorithm of proposition divided using directional images.
In addition, the support vector machines in Statistical Learning Theory is established, the tool relatively powerful as pattern classification, equally
It can also be used for fingerprint classification.And someone develops the algorithm based on support vector machines.Such as:What Yao et al. [2] was proposed
The fingerprint classification method combined using recurrent neural network and support vector machines, Guo et al. [3] propose based on support vector machines
Automatic classification method.One overall thought of these methods is exactly spy of the feature as support vector machines of image of taking the fingerprint
Vector is levied to classify.2004, Gaborl filter and support vector machines were combined and are used for fingerprint point by Batra et al. [4]
Class.They are that each fingerprint image generates 384 dimensional feature vectors using the feature extraction scheme based on Gabor filter.2007
Year, Ji et al. [5] proposes the fingerprint classification method using fingerprint image orientation field and support vector machines combination.Ji passes through calculating
Fingerprint image orientation field has estimated four direction to describe fingerprint image.Each direction percentage is finally calculated, by its structure
Four dimensional vectors are built up to classify to fingerprint as the feature vector of support vector machines.
2007, Jayadeva et al. [6] proposed a kind of non-parallel Hyperplane classification device, referred to as twin support vector machines
(Twin Support Vector Machines, TWSVM) is used for two classification problems.Twin support vector machines is intended to generate two
A not parallel hyperplane so that each hyperplane is closer to one in two classes, and is away as far as possible another class.Two
The also not constraint of parallel condition of a hyperplane finally converts two smaller quadratic programmings of scale for two classification problems and asks
Topic.Obviously, work speed of two smaller quadratic programming problems of scale than solving a large-scale quadratic programming problem is solved
Degree is faster.The time complexity of twin support vector machines is lower, and the training time of algorithm is reduced to the 1/4 of support vector machines, greatly
It improves work efficiency greatly, and in terms of solving with greater advantage, is especially handling especially big data
Classification problem when, superiority is more prominent.Although twin support vector machines processing problem efficiency and aspect of performance all
It is better than traditional support vector machines, but there is still a need for the problem of parameter selection for considering it, i.e. penalty factor and kernel function sheet
The selection of body parameter.2013, Shifei Ding et al. [7] proposed the twin support vector machines based on quantum particle swarm
(Twin Support Vector Machines based on Quantum Particle Swarm Optimizati on,
QPSO-TWSVM).QPSO-TWSVM utilizes ability of the quantum particle swarm in global search, and optimal ginseng is searched in global scope
Number, avoids and falls into locally optimal solution too early, and obtained optimized parameter is more convincing, compared with traditional TWSVM, parameter
Specify more accurate, avoid the blindness of parameter selection, improve the classification accuracy of twin support vector machines.
Summary of the invention
It is an object of the invention to " one-to-many " twin support vector machine classifier of more classifying is applied to fingerprint image point
Class is to improve the reliability of fingerprint recognition.
To achieve the above object, the present invention includes the following steps:
Step 1:Fingerprint image preprocessing:
(1.1) fingerprint image normalizes;
(1.2) segmentation of fingerprint image;
(1.3) binaryzation of fingerprint image;
Step 2:The feature extraction of fingerprint image:
(2.1) fingerprint image gradient fields are sought;
(2.2) fingerprint image orientation field is asked;
(2.3) using all directions percentage be built into an octuple vector as the feature of twin support vector machines to
Amount;
Step 3:" one-to-many " more classify parameter of twin support vector machine classifier is carried out with quanta particle swarm optimization
Optimization;
Step 4:It brings optimized parameter into " one-to-many " to classify in twin support vector machine classifier more, determines final point
Class model simultaneously classifies to fingerprint image.
Detailed description of the invention
Attached drawing 1 is system flow in a kind of fingerprint image categorizing system based on twin support vector machines and method of the invention
Figure.
Attached drawing 2 (a) (b) is respectively a kind of fingerprint image categorizing system and method based on twin support vector machines of the present invention
In original fingerprint image and normalization after fingerprint image.
Attached drawing 3 is after dividing in a kind of fingerprint image categorizing system based on twin support vector machines and method of the invention
Fingerprint image.
After attached drawing 4 is binaryzation in a kind of fingerprint image categorizing system based on twin support vector machines of the invention and method
Fingerprint image.
Attached drawing 5 is that the field of direction is being just in of the invention a kind of fingerprint image categorizing system based on twin support vector machines and method
Friendship is decomposed into coordinate system parameters.
Attached drawing 6 (a) (b) is respectively a kind of fingerprint image categorizing system and method based on twin support vector machines of the present invention
In original fingerprint image and the fingerprint image block directed graph.
Attached drawing 7 is use direction in a kind of fingerprint image categorizing system based on twin support vector machines and method of the invention
Field obtains the example of feature vector.
Attached drawing 8 is QPSO- in a kind of fingerprint image categorizing system based on twin support vector machines and method of the invention
The flow chart of TWSVM algorithm.
Specific embodiment
Step 1:Fingerprint image preprocessing
(1.1) fingerprint image normalizes
Usually collected fingerprint image tonal range changes very greatly, if the tonal range of different fingerprint images all reflected
It is mapped in the same given range, a more unified specification can be provided for subsequent processing, facilitate setting and gray scale
Relevant threshold value.It is the detailed process to fingerprint image normalized below.The reason proposed according to Lin Hong et al. [8]
By, the gray level image I of a fingerprint is quantified as to the matrix of a m × n, wherein I (i, j) indicate pixel (i, j) gray scale
Value, fingerprint image used in the present invention is having a size of 256 × 256, gray level 256.
The gray average and formula of variance of fingerprint image are defined as follows:
Indicate that the gray value after normalizing at pixel (i, j), formula are as follows with G (i, j):
In above formula, M0WithIt is that previously given gray scale desired value and variance (take M in the present invention0=150,
), M and σ2For the average gray and variance of original image.Fingerprint image after being normalized as shown in Fig. 2 (b).
(1.2) segmentation of fingerprint image
The segmentation of fingerprint image is to cut off out from whole image by the fingerprint portion in image, as shown in Figure 3.By
Segmentation to image, has got rid of unnecessary portion, reduces the complexity of calculating and enhances the accurate of feature extraction
Degree.
(1.3) binaryzation of fingerprint image
Fingerprint Image Binarization:Only black and white image is converted gray images into, i.e., black is made by threshold value
The gray value in place is 0, and the gray value in white place is 255.
The present invention has selected the Bernsen algorithm in local thresholding method.The algorithm is one in local binarization algorithm
Good example is a kind of adaptive algorithm of dynamic select threshold value.If gray value of the fingerprint image at pixel (x, y) is
F (x, y), basic thought are:(2w+1) × (2w+1) window in gray level image centered on pixel (x, y) calculates
The threshold value T (x, y) of each pixel (x, y) of image
Fingerprint image is traversed, current grayvalue f (x, y) and present threshold value T (x, y) are compared.
The binary image that b (x, y) is exactly.Fingerprint image after being illustrated in figure 4 binaryzation.
Step 2:The feature extraction of fingerprint image
(2.1) fingerprint image gradient fields are sought
Being illustrated in figure 5 field of direction Orthogonal Decomposition is coordinate system parameters.
1. seeking Z=f (x, y) partial derivative of point (x, y)To acquire gradient vector
2. remembering Gx(x, y) is (x, y) this pointRemember Gy(x, y) is (x, y) this point
So:
Gx(x, y)=F (x+1, y)-F (x, y) (7)
Gy(x, y)=F (x, y+1)-F (x, y) (8)
(2.2) fingerprint image orientation field is asked
Each pixel of fingerprint image is not individualism, and the pixel on it and periphery is associated.It examines
Consider this point, we will be subject to perfect when calculating the direction of certain point by the information of each point around.Specific practice is such as
Under:
If:
So field of direction size is
Wherein, Gx≠ 0 and Gy≠ 0, the partial derivative of (x, y) is found out using sobel operator.Work as GxOr GyWhen being 0, enable θ (i,
J)=0.
In order to construct the feature vector of twin support vector machine classifier, it would be desirable to estimate the field of direction of fingerprint image,
Then statistics calculates field of direction all directions percentage as characteristic value.The present invention estimates the block directed graph of fingerprint using gradient method,
And the field of direction of fingerprint image is indicated using eight directions (π/8 0, π/8, π/4,3, pi/2, π/8 the π/4,7 of 5 π/8,3).
Gradient method calculate block directed graph basic thought be:The each pixel x of fingerprint image is estimated with gradient operator first
Gradient component on direction and the direction y, then fingerprint image is divided into fritter that is equal in magnitude and not overlapping again, according to formula (9)
(10) (11) obtain each piece of mean direction, and as the direction of the block.
Algorithm steps are as follows:
1) using 3 × 3 sobel operator calculate the pretreated each pixel of fingerprint image in the horizontal direction and
Gradient direction on vertical direction:Gx(x, y) and Gy(x,y).Sobel operator representation is as follows:
The direction x:
The direction y:
Gx(x, y)=sobelx*Z (12)
Gy(x, y)=sobely*Z (13)
2) fingerprint image having a size of 256 × 256 is divided into 32 × 32 fritters not overlapped, according to formula (9) (10)
(11), direction of the mean direction as the block of each sub-block is calculated;
3) 2) direction in is changed to angle system by Circular measure, and according to the affiliated range in each piece of direction, it is final to determine
Eight directions:(0 °, 22.5 °, 45 °, 67.5 °, 90 °, 112.5 °, 135 °, 157.5 °) and the block side for drawing out fingerprint image
Xiang Tu, as shown in Fig. 6 (b);
(2.3) all directions percentage is built into an octuple vector as the feature vector of twin support vector machines
Count the number N of the sub-block all directions of θ ≠ 0 in 32 × 32 sub-blocksd, and calculate each direction proportion:
Wherein d ∈ (0 °, 22.5 °, 45 °, 67.5 °, 90 °, 112.5 °, 135 °, 157.5 °),
It combines these direction percentages to obtain the feature vector for being used for supervised training and twin support vector cassification.
As shown in Figure 7.
Step 3:Quanta particle swarm optimization carries out " one-to-many " parameter in twin support vector machine classifier of more classifying
Optimization
Fingerprint image is divided into four classes (left dustpan type, right dustpan type, bucket using more classification methods of " one-to-many " by the present invention
Type, arcuate).In an experiment, there are five parameters altogether for the training function of non-linear twin support vector machines:Four penalty parameter cs 1,
C2, c3, c4 respectively correspond four classifications of fingerprint image, along with the parameter σ of a gaussian kernel function.
Quanta particle swarm optimization is exactly to be searched for by quantum particle swarm optimization to the core that parameter optimizes
The optimized parameter of twin support vector machines.In quanta particle swarm optimization, the position coordinates of particle individual represent it is twin support to
Parameter in amount machine.The particle in population with optimal adaptation angle value is found, the coordinate of the particle is exactly that we need most
Excellent parameter.Finally determine that " one-to-many " classifies twin supporting vector machine model more with the optimized parameter that searches.
The particle that the algorithm represents potential problems solution by M forms a group X={ X1,X2,...,XM, all grains
Son all determines its fitness value by the same objective function.In QPSO-TWSVM algorithm, objective function is exactly to calculate twin branch
The function (classification accuracy is bigger, and the parameter of selection is better) for holding vector machine accuracy rate, if the target search space of problem
It is N-dimensional, then in t moment, the position of i-th of particle is represented by Xi(t)=[Xi,1,Xi,2,...,Xi,N], i=1,2 ...,
M, and particle does not have speed.Particle individual optimum position is expressed as Pi(t)=[Pi,1,Pi,2,...,Pi,N], i=1,2 ..., m,
The global optimum position of particle group is expressed as Gi(t)=[G1,G2,...,GN], and G (t)=Pg(t), wherein g is in the overall situation
The subscript of optimum position particle, g ∈ { 1,2 ..., M }.
Population overall situation best coordinates are represented by:
G (t)=Pg(t) (16)
The renewal equation of QPSO-TWSVM algorithm particle individual location information in each iteration is:
Xi,j(t+1)=Pi,j(t)±α|Cj(t)-Xi,j(t)|·ln[1/ui,j(t)]ui,j~U (0,1) (18)
Wherein,
QPSO-TWSVM finds particle global optimum position by iteration, when iteration to the limit when, be in it is global most
The position coordinates G of the particle of excellent positioni(t)=[G1,G2,...,GN] it is exactly the parameter that we need.QPSO-TWSVM algorithm
Flow chart is as shown in Figure 8.
In this experiment, fingerprint image is from the fingerprint image data library that Institute of Automation Research of CAS provides
FX3000 word bank.Because what this experiment was mainly studied is fingerprint classification, the fingerprint for therefrom picking 513 pieces of clean marks is made
For research object.Wherein 125 pieces of left dustpan type, 147 pieces of right dustpan type, 98 pieces of arcuate, amount to 513 pieces of fingerprints by 143 pieces of bucket type.?
In QPSO-TWSVM algorithm, population scale M=30;Particle position dimension (i.e. classify in twin support vector machines more by " one-to-many "
Number of parameters):N=5;Inertial factor inertia=0.5;Itself factor selfw=2.0;Global factor globalw=
2.0;Mutation probability mutatep=0.05;Iteration maximum times maxgen=100.Using the twin branch of more classification of " one-to-many "
The training function for holding vector machine is trained 70% finger print data, obtains optimized parameter.
Step 4:It brings optimized parameter into " one-to-many " to classify in twin support vector machine classifier more, determines final point
Class model simultaneously classifies to fingerprint image
By optimizing to parameter, obtaining optimized parameter is:C1=511.9121;C2=443.9878;C3=
495.3080;C4=342.1800;S=0.2631 (σ in s, that is, gaussian kernel function).
Optimized parameter is brought into the anticipation function of the twin support vector machines of more classification of " one-to-many " to residue 30%
Finger print data carries out classification prediction.
Prediction result is as shown in table 1.
Classification error rate is:ER=(5+5+3+6)/156=12.18%
Classification accuracy is:AR=1-ER=87.82%
Interpretation of result:
Twin support vector machines is applied to fingerprint classification and obtains higher classification accuracy by this Success in Experiment.
Bibliography
[1]Cappelli R,LuminiA,Maio D,et al.Fingerprint Classification by
Directional Image Partitioning[J].PatternAnalysis&Machine Intelligence IEEE
Transactions on,1999,21(5):402-421.
[2]YaoY,Marcialis GL,Pontil M,et al.Combining flat and
structuredrepresentations for ingerprint classification with recursive neural
networks and supportvectormachines[J].Pattern Recognition,2003,36(2):397-406.
[3]Guo L,WuY,Wu Q,et al.Research onAutomatic Fingerprint
Classification Based on SupportVectorMachine[C]//Intelligent Control
andAutomation,2006.WCICA2006.The Sixth World Congress on.IEEE,2006:4093-4096.
[4]BatraD,Singhal G,Chaudhury S.Gabor filterbasedfingerprint
classification using supportvector machines[C]//India Conference,
2004.Proceedings ofthe IEEE Indicon.IEEE,2004:256-261.
[5]Ji L,YiZ.SVM-basedFingerprint Classification Using Orientation
Field[C]//International Conference onNatural Computation.IEEE,2007:724-727.
[6]Jayadeva,Khemchandani R,Chandra S.Twin SupportVectorMachines
forPattern Classification[J].IEEE Transactions on PatternAnalysis&Machine
Intelligence,2007,29(5):905-910.
[7]Shifei Ding,FulinWu,Nie Ru et al.Twin SupportVector Machines Based
on Quantum Particle Swarm Optimization[J].JOURNAL OF COMPUTERS,2013,8(7):
1743-1750.
[8]Hong L,WanY,JainA.Fingerprint Image Enhancement:Algorithm
andPerformance Evaluation[M].IEEE Computer Society,1998.
Claims (7)
1. a kind of fingerprint image categorizing system and method based on twin support vector machines, which is characterized in that mainly include:
Step 1:Fingerprint image preprocessing:
(1.1) fingerprint image normalizes;
(1.2) segmentation of fingerprint image;
(1.3) binaryzation of fingerprint image;
Step 2:The feature extraction of fingerprint image:
(2.1) fingerprint image gradient fields are sought;
(2.2) fingerprint image orientation field is asked;
(2.3) all directions percentage is built into an octuple vector as the feature vector of twin support vector machines;
Step 3:Quanta particle swarm optimization optimizes " one-to-many " parameter in twin support vector machine classifier of more classifying;
Step 4:It brings optimized parameter into " one-to-many " to classify in twin support vector machine classifier more, determines final classification mould
Type simultaneously classifies to fingerprint image.
2. a kind of fingerprint image categorizing system and method based on twin support vector machines according to claim 1, feature
It is, in step 1.1, fingerprint image is having a size of 256 × 256, gray level 256, in normalized, M0WithIt is pre-
First given gray scale desired value and variance (take M in experiment0=150,)。
3. a kind of fingerprint image categorizing system and method based on twin support vector machines according to claim 1, feature
It is, cuts off out from whole image by the fingerprint portion in image in step 1.2, gets rid of unnecessary portion, from
And reduce the accuracy of the complexity calculated and Enhanced feature extraction.
4. a kind of fingerprint image categorizing system and method based on twin support vector machines according to claim 1, feature
It is, when binary conversion treatment in step 1.3 has selected the Bernsen algorithm in local thresholding method.
5. a kind of fingerprint image categorizing system and method based on twin support vector machines according to claim 1, feature
It is, in step 2.1, the gradient vector calculation formula of fingerprint image picture point (x, y) is
Gx(x, y)=F (x+1, y)-F (x, y)
Gy(x, y)=F (x, y+1)-F (x, y)
6. a kind of fingerprint image categorizing system and method based on twin support vector machines according to claim 1, feature
Be, in step 2.2, using the block directed graph of fingerprint image gradient method estimation fingerprint, and using eight directions (0, π/8, π/
4,3 π/8, pi/2,5 π/8,3 π/4,7 π/8) indicate the field of direction of fingerprint image, calculate all directions percentage and by this
A little direction percentage combinations are to obtain the feature vector for supervised training and twin support vector cassification.
7. a kind of fingerprint image categorizing system and method based on twin support vector machines according to claim 1, feature
It is, in step 3, fingerprint image is divided into four classes:Left dustpan type, right dustpan type, bucket type, arcuate.In an experiment, non-linear twin
There are five parameters altogether for the training function of support vector machines:Four penalty parameter cs 1, c2, c3, the ginseng of c4 and gaussian kernel function
Number σ, environmentally optimizes the parameter of twin support vector machines in Matlab using quanta particle swarm optimization, in this experiment
In, fingerprint image is the fingerprint image data library FX3000 word bank provided from Institute of Automation Research of CAS.Because this
What experiment was mainly studied is fingerprint classification, so therefrom picking the fingerprint of 513 pieces of clean marks as research object.It is wherein left
125 pieces of dustpan type, 147 pieces of right dustpan type, 98 pieces of arcuate, amounts to 513 pieces of fingerprints by 143 pieces of bucket type.In QPSO-TWSVM algorithm, kind
Group's scale M=30;Particle position dimension (i.e. " one-to-many " more classify the number of parameters in twin support vector machine classifier):N
=5;Inertial factor inertia=0.5;Itself factor selfw=2.0;Global factor globalw=2.0;Mutation probability
Mutatep=0.05;Iteration maximum times maxgen=100.In experiment, using the twin supporting vector of more classification of " one-to-many "
The training function of machine is trained 70% finger print data, obtains optimized parameter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810429965.0A CN108921186A (en) | 2018-05-08 | 2018-05-08 | A kind of fingerprint image categorizing system and method based on twin support vector machines |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810429965.0A CN108921186A (en) | 2018-05-08 | 2018-05-08 | A kind of fingerprint image categorizing system and method based on twin support vector machines |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108921186A true CN108921186A (en) | 2018-11-30 |
Family
ID=64402816
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810429965.0A Pending CN108921186A (en) | 2018-05-08 | 2018-05-08 | A kind of fingerprint image categorizing system and method based on twin support vector machines |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108921186A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109919202A (en) * | 2019-02-18 | 2019-06-21 | 新华三技术有限公司合肥分公司 | Disaggregated model training method and device |
CN111368920A (en) * | 2020-03-05 | 2020-07-03 | 中南大学 | Quantum twin neural network-based binary classification method and face recognition method thereof |
CN112101213A (en) * | 2020-09-15 | 2020-12-18 | 哈尔滨理工大学 | Method for acquiring fingerprint direction information |
CN112773365A (en) * | 2019-10-22 | 2021-05-11 | 上海交通大学 | System for monitoring mental load of underwater vehicle during underwater operation |
-
2018
- 2018-05-08 CN CN201810429965.0A patent/CN108921186A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109919202A (en) * | 2019-02-18 | 2019-06-21 | 新华三技术有限公司合肥分公司 | Disaggregated model training method and device |
CN112773365A (en) * | 2019-10-22 | 2021-05-11 | 上海交通大学 | System for monitoring mental load of underwater vehicle during underwater operation |
CN111368920A (en) * | 2020-03-05 | 2020-07-03 | 中南大学 | Quantum twin neural network-based binary classification method and face recognition method thereof |
CN111368920B (en) * | 2020-03-05 | 2024-03-05 | 中南大学 | Quantum twin neural network-based classification method and face recognition method thereof |
CN112101213A (en) * | 2020-09-15 | 2020-12-18 | 哈尔滨理工大学 | Method for acquiring fingerprint direction information |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112418117B (en) | Small target detection method based on unmanned aerial vehicle image | |
Pramanik et al. | Granulated RCNN and multi-class deep sort for multi-object detection and tracking | |
CN108921186A (en) | A kind of fingerprint image categorizing system and method based on twin support vector machines | |
CN108256456B (en) | Finger vein identification method based on multi-feature threshold fusion | |
CN111079674B (en) | Target detection method based on global and local information fusion | |
CN104537673B (en) | Infrared Image Segmentation based on multi thresholds and adaptive fuzzy clustering | |
Katiyar et al. | A hybrid recognition system for off-line handwritten characters | |
Tuba et al. | Handwritten digit recognition by support vector machine optimized by bat algorithm | |
CN101140623A (en) | Video frequency objects recognition method and system based on supporting vectors machine | |
CN109325507B (en) | Image classification method and system combining super-pixel saliency features and HOG features | |
CN104778457A (en) | Video face identification algorithm on basis of multi-instance learning | |
CN113408605A (en) | Hyperspectral image semi-supervised classification method based on small sample learning | |
Zhu et al. | Semi-supervised streaming learning with emerging new labels | |
Badawi et al. | A hybrid memetic algorithm (genetic algorithm and great deluge local search) with back-propagation classifier for fish recognition | |
CN107944356B (en) | The identity identifying method of the hierarchical subject model palmprint image identification of comprehensive polymorphic type feature | |
CN105825233B (en) | A kind of pedestrian detection method based on on-line study random fern classifier | |
CN110826408B (en) | Face recognition method by regional feature extraction | |
CN110532946A (en) | A method of the green vehicle spindle-type that is open to traffic is identified based on convolutional neural networks | |
CN106204651A (en) | A kind of method for tracking target based on the judgement improved with generation conjunctive model | |
CN104616319A (en) | Multi-feature selection target tracking method based on support vector machine | |
CN104881871A (en) | Traffic image segmentation method based on improved multi-object harmony search algorithm | |
Mittal et al. | Fake-face image classification using improved quantum-inspired evolutionary-based feature selection method | |
Shi et al. | Face recognition algorithm based on self-adaptive blocking local binary pattern | |
CN102081740B (en) | 3D image classification method based on scale invariant features | |
Serdouk et al. | An improved artificial immune recognition system for off-line handwritten signature verification |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181130 |
|
WD01 | Invention patent application deemed withdrawn after publication |