CN107909004A - A kind of 3D palmprint recognition technologies - Google Patents
A kind of 3D palmprint recognition technologies Download PDFInfo
- Publication number
- CN107909004A CN107909004A CN201710996196.8A CN201710996196A CN107909004A CN 107909004 A CN107909004 A CN 107909004A CN 201710996196 A CN201710996196 A CN 201710996196A CN 107909004 A CN107909004 A CN 107909004A
- Authority
- CN
- China
- Prior art keywords
- palmmprint
- feature
- recognition
- identification
- rarefaction representation
- 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
- 238000005516 engineering process Methods 0.000 title claims abstract description 40
- 238000000605 extraction Methods 0.000 claims abstract description 18
- 230000010354 integration Effects 0.000 claims abstract description 5
- 230000004927 fusion Effects 0.000 claims description 10
- 238000000034 method Methods 0.000 abstract description 85
- 230000000694 effects Effects 0.000 abstract description 10
- 238000004364 calculation method Methods 0.000 abstract description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000012549 training Methods 0.000 description 18
- 238000012360 testing method Methods 0.000 description 17
- 239000013598 vector Substances 0.000 description 13
- 238000011160 research Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 8
- 239000011159 matrix material Substances 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 238000005286 illumination Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 210000001525 retina Anatomy 0.000 description 5
- 230000009466 transformation Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 230000003542 behavioural effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 230000004069 differentiation Effects 0.000 description 3
- 238000012880 independent component analysis Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000011840 criminal investigation Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000003708 edge detection Methods 0.000 description 2
- 210000000887 face Anatomy 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 210000003491 skin Anatomy 0.000 description 2
- 206010011409 Cross infection Diseases 0.000 description 1
- 244000060701 Kaempferia pandurata Species 0.000 description 1
- 206010029803 Nosocomial infection Diseases 0.000 description 1
- 235000016390 Uvaria chamae Nutrition 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 229910002056 binary alloy Inorganic materials 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 210000002615 epidermis Anatomy 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 210000003195 fascia Anatomy 0.000 description 1
- 230000005021 gait Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000033001 locomotion Effects 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000001568 sexual effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000037303 wrinkles Effects 0.000 description 1
- 210000000707 wrist Anatomy 0.000 description 1
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional objects
-
- 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)
- Multimedia (AREA)
- Data Mining & Analysis (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 3D palmprint recognition technologies, two parts are included：The multi-feature extraction and integration technology of palmmprint and the palmprint recognition technology based on the classification of neighborhood rarefaction representation, the beneficial effects of the invention are as follows：1st, for image and three-dimensional palm print data, multidimensional characteristic is extracted, plays different Features Complements.The shortcomings that overcoming single, one-dimensional feature, plays the comprehensive and integrality that multiple features describe information.2nd, a kind of improved rarefaction representation sorting technique is studied --- neighborhood rarefaction representation is classified（NSRC）, the Classification and Identification of palmmprint is completed using NSRC, had not only ensured the recognition effect of original rarefaction representation sorting technique, but also certain calculation amount can be reduced.
Description
Technical field
The present invention relates to a kind of characteristics of human body's identification technology, is specifically a kind of 3D palmprint recognition technologies.
Background technology
In the modern society of advanced IT application, with traffic, communication, network technology high speed development, information security shows
Go out unprecedented importance.In daily life and many occasions such as finance, the administration of justice, safety check, e-commerce need it is accurate
Identification to ensure the safety of system, therefore, the application of the identity recognizing technology of people is more and more important.Traditional identity mirror
Two classes can be generally divided into by determining means：Based on specific having, such as identity card, credit card, key, employee's card；It is based on
Specific knowledge, such as password, password, code word.Many occasions need both approaches being combined, such as withdraw the money in ATM machine
When, you not only need credit card, it is also necessary to know password.The shortcomings that conventional method, first consists in specific having or knowledge may
Forget, lose or be stolen.There is the difficulty in memory using specific knowledge, people are difficult to remember complicated password, and simple close
Code and is easy to be hypothesized or decodes (such as birthday, telephone number).Secondly, it does not bind uniquely with user, once others
These havings or knowledge are obtained, he will possess the right same with the owner of lost property.The development and raising of biometrics identification technology, are
The certification and identification of identity provide more preferable means.
So-called living things feature recognition just refers to using human body intrinsic physiological characteristic or behavioural characteristic, to carry out a person
Part identification.Physiological characteristic refers to human body parts are carried out directly to measure obtained data, some represent the physiological characteristic bag of type
Include fingerprint, face, iris, retina, palm shape etc.；Behavioural characteristic is the measurement to personal habits sexual act, is to characteristics of human body
Indirect measurement, some representational behavioural characteristics include sound ripple, keystroke custom, signature etc..Physiological characteristic and behavior
The differentiation of feature is manually given, and in many occasions, boundary therebetween may not be very stringent.Most common biology is special
Levying identification technology includes fingerprint, face, sound ripple, iris, retina, palm shape, signature, palmmprint etc..Also some technologies because
For use it is inconvenient or immature lack commercial value, these technologies include DNA, ear type, smell, gait, finger-type, finger
First texture etc..
Although palmprint image is applied to criminal investigation field already, automatic personal recognition is then as a kind of biological identification technology
It is developed in recent years.Its sampling can facilitate as fingerprint and hand, and can ensure stability and uniqueness, together
When, personal recognition can't allow people to associate the infringement of the right of privacy, the acceptable degree of its testee is higher, and identify system
Standardization level of the Hardware standardization degree of system also than recognition of face, speech recognition, retina identification is high.Discrimination and identification
Speed can reach the requirement of utility system.Therefore the identity recognizing technology based on palmmprint is the weight of human body identity recognizing technology
Want content.And palmprint image is low contrast, the image of low resolution, and the feature in this sampled images is extracted in image procossing
It is highly difficult.Thus the research to palmprint recognition technology has important reality and theory significance.
Personal recognition is an emerging biometrics identification technology with stronger market prospects.Contain what is enriched in palmmprint
Feature, such as main line, gauffer, crestal line and minutiae point.Above-mentioned whole features can be extracted in high-resolution palm print image,
And in low resolution palmprint image, main line and gauffer feature can only be extracted, high-resolution palm print identification technology is mainly applied
In fields such as criminal investigations.Since the acquisition of high-resolution palm print image is relatively difficult and picture size is big etc., reason, correlation are ground
Study carefully not development extensively.And low resolution palmprint recognition technology has that image acquisition is easy, processing speed is fast and accuracy of identification
The advantages that high, be adapted to generally civilian and commercial, is the emphasis of academic circles at present research.
Palm be finger root to the region between wrist, the dermatoglyph of palm is palmmprint.Palmmprint mainly extracts main line
With gauffer feature.(1) main line：Since deep fascia is connected with skin of palm of hand surface, the broken line of epidermis infolding is when palm is held inside
Fixed, formed with the opening and closing during palm grip.Because the mode that different people grasps, the size of palm, palm
The difference of thickness, causes main line also to have obvious difference.Main line is feature the most obvious in personal recognition.(2) gauffer：Pleat
Wrinkle is many irregular shallower, thinner curves or straight line in palmmprint.Some folds are that day is born with and is had, after some folds are
What it was formed, due to some position proper motions, caused skin is stretched or shunk and loses bullet in different directions
Property formed permanent fold.
Emerging technology of the personal recognition as biometrics identification technology, has unique line feature, in low resolution figure
Its metastable feature can also be extracted as in, there is stronger anti-noise ability.With other common living things feature recognitions
Technology is compared, and personal recognition has many unique characteristics.
1) compared with recognition of face, palm print characteristics from the factors such as jewelry, expression and posture influence, relatively surely
It is fixed, it can preferably ensure the accuracy of identification of Palm Print Recognition System.
2) compared with fingerprint technique, the effective coverage of palmmprint will be far longer than fingerprint, and palmmprint contains more more rich than fingerprint
Information.And fingerprint is only applicable to the preferable crowd of fingerprint quality, limit that fingerprint is unintelligible or fingerprint is seriously ground
The people of damage.In addition, palmmprint collecting device cost is lower than collection fingerprint, because fingerprint collecting needs high-precision equipment.
3) compared with iris, retina identification, although iris, retina discrimination are higher, collecting device is high,
And cause eye illness cross infection, and palmmprint collecting device cost is relatively low, image acquisition mode is simple, and acceptable degree is high.
Due to the projective transformation in image acquisition procedures, the loss of many characteristic informations is caused, causes two-dimentional knowledge
Do not influenced by illumination and posture huge.In order to improve discrimination, influence of the change of visual angle and illumination to discrimination is avoided, also
Living things feature recognition is as set further to be perceived naturally close to the mankind, numerous researchers start to be directed to the knowledge of three dimensional biological feature
Not.3D data truly reflect the shape of 3D objects in space.Projective transformation need not be considered with 3D data, due to the figure of two dimension
As data are substantially projection of the 3D objects on two-dimensional space, cause same target that there is changeable two dimension on a projection plane
Performance, i.e., the diversity that the palmprint image of the same palm changes with posture, this is also the personal recognition for being currently based on 2D images
One of greatest difficulty that method faces.
For the palmprint image of low resolution, main line and gauffer information realization personal recognition are mainly utilized.According to palmmprint
The expression of middle feature and matching process, substantially can be divided into four classifications by palm grain identification method, be structure-based side respectively
Method, Statistics-Based Method, the method based on subspace and the method based on coding.In addition, also list some is not belonging to this
A little class method for distinguishing.
Structure-based method
Structure-based method is primarily referred to as direction using main line in palmmprint and gauffer and positional information realizes that palmmprint is known
Method for distinguishing.This kind of method is mainly made of two parts：1) streakline feature in palmmprint is extracted；2) streakline feature is effective
Represent and match.The effective of line feature represents to be primarily referred to as easy to matching, and takes memory space as few as possible.For spy
The extraction of sign, it is more to use various line detections to calculate.Son and edge detection operator；Expression for feature, is mainly adopted
Palmmprint streakline is replaced with straightway or characteristic point；And the matching of feature uses Euclidean distance between characteristic point etc. mostly.
Wu etc. proposes a kind of method of the derivative extraction palmmprint major line features using Gaussian function.This method uses four
Different detective operators, detect the streakline of four different directions respectively, finally merge the testing result of all directions.In order to overcome
The influence that non-linear deformation and rotation are brought, the bianry image of expression streakline is rotated before matching for document [2] and form
Learn expansion, afterwards again with other palmprint match.Obtain the line feature of palmmprint in document [3] by binaryzation first, extract afterwards
Some longest streaklines of span in palmmprint inside maximum inscribed circle, using similar Hausdorff distances during matching
Bi-directional matching method.
Structure-based method is the method for personal recognition of early stage.Generally speaking, it is mostly based on the side of structure
Method is all reference or transplants from the method in fingerprint recognition, simple, intuitive.But this kind of method straightway or characteristic point are approximate
Ground represents palmmprint streakline, lost bulk information, thus the not high of discrimination in addition, the recognition performance of this kind of method in very great Cheng
Edge detection operator or wide line detector are depended on degree.The relatively fine fuzzy line of some in palmmprint includes substantial amounts of differentiation
Information, but operator can not be detected and detected.Substantial amounts of straightway and characteristic point also make matching process very time-consuming.
Statistics-Based Method
Statistics-Based Method refers to the identification by the use of statistics such as the centers of gravity, average, variance of palmprint image as feature
Method, can be further divided into methods of the method based on local statistic and global statistics wherein based on local statistic needs
Some fritters are divided the image into, count the statistical informations such as every piece of average and variance afterwards, are finally connected as representing entirely slapping
The feature vector of line；And the method based on global statistics then directly calculates the statistical information conducts such as square and the center of gravity of whole image
During the feature matchings of palmmprint, generally related coefficient, single order norm or the Euclidean distance of vector more often are used
For the method based on local statistic, before statistical nature is extracted, it usually needs image is converted, such as
.Li such as Fourier transformation, wavelet transformation etc. [4] extract the frequency domain information of palmprint image with Fourier transformation first, afterwards piecemeal
And calculate every piece amplitude and phase and have the accurate of higher as the structure-based method of feature Statistics-Based Method ratios
Rate, the essence of statistics is also so that such method is to insensitive for noise
Method based on subspace
Subspace method regards palmprint image as high dimension vector or matrix, by projecting or converting, is translated into low
Dimensional vector or matrix, and palmmprint is represented and matched under this lower dimensional space.Mainly include independent component analysis
(Independent component analysis, ICA), principal component analysis (Principal component analysis,
PCA), linear discriminant analysis (Linear discriminant analysis, LDA) etc..Method based on subspace mostly needs
Training set is constructed to the palmprint image of each classification, optimal projection vector or matrix be calculated on the training set, and will throw
Feature of the vector or matrix of movie queen as such palmmprint.In cognitive phase, palmprint image to be measured is made first same projection or
Conversion, is classified using nearest neighbor classifier afterwards.Method based on subspace has been applied successfully to recognition of face, is transplanted to the palm
Good effect is also achieved after line identification.
Lu etc. proposes the Eigen-Palm methods that dimensionality reduction is carried out using PCA, and palmprint image is connected as by this method first
High dimension vector, and the characteristic value and feature vector of the vectorial Scatter Matrix are calculated, retain some larger characteristic values pair afterwards
The feature vector answered forms projection matrix.Due to PCA primary concern is that the expression (Representation) of palmmprint, rather than
The differentiation (Discriminance) of palmmprint, Wu etc. have also been proposed the FisherPalm for carrying out LDA dimensionality reductions again on the basis of PCA
Method.LDA methods consider divergence and class scatter in class at the same time, and minimize divergence in class at the same time by maximizing class scatter
Calculate optimal projection matrix.
Method based on subspace has firm theoretical foundation, and has been widely used in recognition of face.Relatively
In structure-based method, have the advantages that discrimination is high, feature is small, also there is the identification of higher than Statistics-Based Method
Rate.
Method based on coding
Method based on coding refers to first filter palmprint image with wave filter, afterwards will be filtered according to some rules
As a result the method encoded.Usual feature stores all in accordance with the form of bit code, and binary system is used to obtained condition code more
"AND" or distance calculate similarity.Such method mainly includes three cores:Selection (Gabor, the height of wave filter
This, the derivative of Gauss), coding rule (maximum, ordinal number relation), and matching way (point-to-point, to put to region).
Kong etc. propose it is a kind of be known as FusionCode method with the Gabor filter of both direction to palmmprint
After image filtering, to each sampled point, the phase information in the direction of coded magnitude maximum.Phase information is quantized to four areas
Between, therefore each sampled point is equally encoded as 2 bits.FusionCode methods make use of amplitude and phase information at the same time,
Therefore more preferable recognition performance is obtained.The it is proposeds such as Wu are by the use of the derivative of Gaussian function as wave filter, in horizontal and vertical side
To image filtering, finally according to the symbolic coding of filter result, it is referred to as DoGCode. test result indicates that this method is better than
FusionCode.Kong etc. proposes to filter palmprint image using the real value Gabor filter of six direction in the literature, and right
The direction encoding of amplitude minimum, is known as Competition coding (Competitive code).Dexterously it is encoded in each sampled point
After three bits, the angular distance between sampled point can be efficiently calculated by binary exclusive-OR operation.In order to gram
The influence that the clothes translation that brings of pretreatment produces matching, matching stage need to carry out in the range of [- 2 ,+2] vertically and
Translated in horizontal direction, and take the minimum matching result of distance as final distance since Competition coding has investigated palmmprint figure
The directional information of picture, it is insensitive to intensity of illumination, and have good stability to the palmmprint of different times collection, therefore obtain
Very high accuracy of identification.
The method that method based on coding has initially used for reference iris recognition, afterwards researcher proposed according to the characteristics of palmmprint
Some improvement projects.From amplitude and phase information is utilized, to the directional information of extraction palmmprint；From single filter, to use
One group of wave filter (Filter bank) of different directions, test result indicates that these improvement are feasible and effective.
Sorting technique based on rarefaction representation
This sorting technique is applied in recognition of face first by Wright etc., obtains excellent effect, is presented it and is being divided
Superiority in class identification problem.Height et al. constrains core and adds sparse representation model, it is proposed that nuclear sparse expression recognition of face
Method；Yang et al. combines Scale-space theory and rarefaction representation, it is proposed that pyramidal sparse based on linear space
Represent recognition methods.Guo et al. [14] carries out the identification of two-dimentional palmmprint using rarefaction representation sorting technique, but the feature used
Only two-dimentional palmmprint monochrome information.
Method based on three-dimensional palm print data
Although the research of current 3D objects is also fewer, some researchs of scientific research group to 3D faces have been achieved for one
Fixed achievement, for example：Lu carries out three-dimensional face identification using ICP methods.This method assumes that the 3-D view in picture library is one
A more complete faceform, search 3-D view are a front views just as picture library image.2008, D.Zhang etc.
An article using 3D personal recognitions is delivered, this is also the article that an international first piece introduces 3D personal recognitions.In addition, send out recently
One of table, it is exactly the algorithm based on curvature using Differential Geometry that used recognizer is mentioned in article, is obtained
Good recognition effect.Yang represents to go the geometrical property of the regional area of description three-dimensional palm print, extraction using Shape Indexes
Gabor wavelet feature is used to identify palmmprint, obtains better effects.
Palmmprint has a very big potentiality due to some advantages of itself, in field of biological recognition, but personal recognition at present
Main problem is：
(1) the extraction Shortcomings of palm print characteristics：On the premise of illumination and posture is limited, the two dimensional character of acquisition can reach
To preferable recognition effect, but when illumination and posture change, the two dimensional character of extraction can reduce personal recognition effect.
Palmmprint Research on classifying method is relatively fewer：Efficiently palmmprint sorting algorithm compares shortage at present.
The content of the invention
It is an object of the invention to provide a kind of 3D palmprint recognition technologies, to solve proposed in the above background technology ask
Topic.
To achieve the above object, the present invention provides following technical solution：
A kind of 3D palmprint recognition technologies, comprise the steps of：
The extraction and fusion of palmmprint；
The identification of palmmprint.
Further scheme as the present invention：The step A is extracted using multicharacteristic information and integration technology, the feature
Information includes the Gabor wavelet feature of two dimension and three-dimensional curvature of curved surface feature, by the feature of 2-D data and three-dimensional data
Feature obtains multidimensional characteristic by Fusion Features, ensure that the comprehensive and integrality of information.
Further scheme as the present invention：The identification of the palmmprint is realized based on sparse identification technology.
Further scheme as the present invention：The Gabor wavelet feature extraction uses Gabor filter.
Compared with prior art, the beneficial effects of the invention are as follows：1st, it is special for image and three-dimensional palm print data, extraction multidimensional
Sign, plays different Features Complements.The shortcomings that overcoming single, one-dimensional feature, play multiple features information is described it is comprehensive
And integrality.2nd, a kind of improved rarefaction representation sorting technique is studied --- neighborhood rarefaction representation is classified (NSRC), utilizes NSRC
The Classification and Identification of palmmprint is completed, had not only ensured the recognition effect of original rarefaction representation sorting technique, but also certain calculating can be reduced
Amount.
Brief description of the drawings
Fig. 1 is palmmprint automatic identification flow chart.
Fig. 2 is the curvature comparison diagram in Differential Geometry.
Embodiment
The technical solution in the embodiment of the present invention will be clearly and completely described below, it is clear that described implementation
Example is only part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, this area is common
Technical staff's all other embodiments obtained without making creative work, belong to the model that the present invention protects
Enclose.
A kind of 3D palmprint recognition technologies, include two parts：
(1) multi-feature extraction and integration technology of palmmprint
Two-dimentional palmprint image extracts Gabor wavelet feature, three-dimensional palm print data characteristics extraction average curvature feature.Gabor
Wave filter is narrow band filter, has preferable resolution capability in spatial domain and frequency domain, can be reached in spatial domain and frequency domain
It is optimal to combining, there is obvious set direction and frequency selectivity, be widely used in Texture Segmentation and target identification neck
Domain.Gabor filter is suitable for texture feature extraction.
Curvature is a kind of intrinsic attribute of curved surface, the distance dependent of the point in its value and curved surface, with curved surface in space
Position is unrelated.Curvature portrays the degree of crook in certain point, and average curvature can determine the shape of curved surface, and Gaussian curvature is portrayed convex
The shape of the convex domain of curved surface and non-convex surface.We using calculate curvature method be first using 7 × 7 template and image
Convolution is carried out, tries to achieve single order partial differential and second order partial differential, the definition then in conjunction with average curvature and Gaussian curvature obtains curvature
Value.After curvature value is obtained, curvature value is mapped on gray level image and obtains average curvature chart picture and Gaussian curvature image.
According to the difference of level residing for information to be fused, information fusion can be generally divided into data Layer fusion, characteristic layer melts
Conjunction and Decision-level fusion.This project mainly uses Feature-level fusion.The feature of the feature of 2-D data and three-dimensional data is passed through
Fusion Features obtain multidimensional characteristic, ensure that the comprehensive and integrality of information.
(2) palmprint recognition technology based on the classification of neighborhood rarefaction representation
Rarefaction representation is the research hotspot of field of signal processing in recent years, it is a kind of decomposable process to original signal,
Input signal is expressed as the process of the linear approximation of dictionary by a dictionary obtained in advance by the decomposable process.
Although the Optimized model of rarefaction representation is the angle design from signal reconstruction, its represent result have in identification compared with
Good performance.Rarefaction representation be used for identify, i.e., based on sparse identification (SRC), this sorting technique by Wrightet etc. first
Applied in recognition of face, obtain excellent effect, present its superiority in classification and identification.SRC methods are very
More different classes of object is put into training set, when needing the object unknown to some classification to classify, with every in training set
The linear combination of a sample describes the object of this unknown classification.Then calculate residual between reconstructed sample and test sample
Difference, when residual error is a kind of especially small at certain, and other classifications are especially big, the object of the unknown classification belongs to such.But when residual
Difference presentation is more balanced, and very big situation, shows that test object is not belonging to the classification of training set.
The present invention operation principle be：
(1) multi-feature extraction and integration technology of palmmprint
For Complicated Problems of Pattern Recognition, such as Handwritten Chinese Character Recognition, personal recognition, there is presently no a simple side
Method, which can reach higher discrimination and reliability, each method, the advantages of respective, defect and the different scope of application,
There is certain complementarity between different features and recognition methods.Therefore different methods is organically combined to play
Respective advantage, overcomes one's shortcomings, the identifying system of configuration information pattern of fusion, is a Main way of pattern identification research.
For in the feature of personal recognition, two dimensional character to prepare to use Gabor wavelet, and three-dimensional feature is intended using average curvature
Three-dimensional feature.Gabor filter is narrow band filter, has preferable resolution capability in spatial domain and frequency domain, can be with
It is optimal to reach joint in spatial domain and frequency domain, there is obvious set direction and frequency selectivity, is widely used in texture point
Cut and field of target recognition.Gabor filter is suitable for texture feature extraction.
The cosine form of two-dimensional Gabor filter can be defined as the product of Gaussian function and cosine function：
In formula：
(x, y) --- filter center position；
The frequency of f --- cosine signal, i.e. filter centre frequency；
--- the direction of Gabor filter；
σx--- the effective width of Gauss function；
σy--- the effective length of Gauss function.
By varying parameter (σx, σy, f,), various forms of Gabor filters can be obtained.
The feature extraction of three-dimensional palm print model uses curvature of curved surface.Curvature of curved surface can describe the curved surface that three-dimensional palm print enriches
Gauffer information.Curvature is a kind of intrinsic attribute of curved surface, the distance dependent of the point in its value and curved surface, with curved surface in space
Position is unrelated.Curvature portrays the degree of crook in certain point, and average curvature can determine the shape of curved surface, and Gaussian curvature is portrayed convex
The shape of the convex domain of curved surface and non-convex surface.Define the arc length parameter s on curve α so that α (0)=p, α ' (0)=T, T
It is unit tangent vectors of the point p along curve α.By α " (0)=kp(T) n defines normal curvature ks of the curved surface S in point pp(T), n is song
Face S is in the method direction of point p.
As plane ΠpWhen rotating a circle around n, α, T is changed correspondingly, and obtains different normal curvature kp(T).Relative to maximum and
Direction { the T of minimum normal curvature1,T2It is known as the principal direction of curved surface S, corresponding obtained curvature is known as principal curvatures,Gaussian curvature and average curvature are defined as：
They are all inherent geometrical invariants, portray the degree of crook of curved surface well.
(2) palmprint recognition technology based on the classification of neighborhood rarefaction representation
Rarefaction representation is the research hotspot of field of signal processing in recent years, it is a kind of decomposable process to original signal,
Input signal is expressed as the process of the linear approximation of dictionary by a dictionary obtained in advance by the decomposable process.Represent such as
Under：
Y=Ax, wherein y are pending signals, and A is dictionary, and x represents for coefficient vector, | | x | |0For the degree of rarefication of x, table
Show number non-zero in x.Under the premise of one group of dictionary A is given, the purpose of sparse statement is to make x as far as possible sparse, reconstruction signal y：
min||x||0
S.t.y=Ax
Although the Optimized model of rarefaction representation is the angle design from signal reconstruction, its represent result have in identification compared with
Good performance.Rarefaction representation is used to identify, i.e., based on sparse identification
(SparserepresentationBasedClassification, SRC).Wright et al. utilizes the sparse table of facial image
Show carry out recognition of face, obtain preferable effect.This method assumes that a given test sample can use all training samples
Sparse linear combination carry out approximate representation.In linear combination, probability that the training sample similar with test sample is not zero
It is very big, and the coefficient of other training samples is generally zero or near zero.
In SRC, solving sparse nonzero coefficient can be obtained by solving L0 norm optimization problems.But solve L0
Norm optimization problem is a NP problem.According to sparse representation theory, in some cases, L0 norm optimizations problem is looked for most
Excellent solution is equivalent to look for the optimal solution of L1 norm problems.So in SRC, can be with L1 norm optimizations come excellent instead of L0 norms
Change.
SRC methods do classify when, it is necessary to test sample be represented with whole training samples, when training sample is very big
When, the calculation amount of this method can be very big.The improved SRC methods of this subject study, enabling reduce this calculation amount.Using elder generation
A part of training sample most like with test sample is obtained by certain mode, is referred to as the N neighborhoods training sample of the test sample
This, then carries out rarefaction representation to test sample with obtained N neighborhoods training sample, finally judges the classification of test sample.
For arbitrary test sample Y, improved SRC algorithms are first found and the most similar N neighborhoods training sample sets of test sample Y
Close.Assuming that find n in the i-th class training sampleiA training sample, then the i-th class training sample be denoted asObtained N neighborhood training samples are denoted as A={ A1,A2,…As}∈RM×N.Then, solve
Following optimization problem：
X is rarefaction representation coefficients of the Y by A column vectors.According to this coefficient, the reconstructed sample of calculating test sample YWhereinRepresent only to retain in rarefaction representation coefficient
With the vector obtained after the relevant coefficient of the i-th class.Finally, test sample Y and reconstructed sample Y is calculatediBetween residual errorResidual error is set to reach the classification that minimum i is exactly test sample.
It is as follows based on rarefaction representation sorting algorithm step：
(1) dictionary A is determined
(2) Y to be measured is encoded by dictionary A, makes its coefficient L1 Norm minimums, λ is regular parameter
(3) residual error is calculated：
Classify to sample to be tested：Identity (y)=argmin (ri)。
Claims (4)
1. a kind of 3D palmprint recognition technologies, it is characterised in that comprise the steps of：
The extraction and fusion of palmmprint；
The identification of palmmprint.
2. 3D palmprint recognition technologies according to claim 1, it is characterised in that the step A is carried using multicharacteristic information
Take and integration technology, the characteristic information includes the Gabor wavelet feature of two dimension and three-dimensional curvature of curved surface feature, by two-dimemsional number
According to feature and the feature of three-dimensional data obtain multidimensional characteristic by Fusion Features, ensure that the comprehensive and integrality of information.
3. 3D palmprint recognition technologies according to claim 1, it is characterised in that the identification of the palmmprint is based on sparse
What identification technology was realized.
4. 3D palmprint recognition technologies according to claim 2, it is characterised in that the Gabor wavelet feature extraction uses
Gabor filter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710996196.8A CN107909004A (en) | 2017-10-23 | 2017-10-23 | A kind of 3D palmprint recognition technologies |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710996196.8A CN107909004A (en) | 2017-10-23 | 2017-10-23 | A kind of 3D palmprint recognition technologies |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107909004A true CN107909004A (en) | 2018-04-13 |
Family
ID=61840703
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710996196.8A Pending CN107909004A (en) | 2017-10-23 | 2017-10-23 | A kind of 3D palmprint recognition technologies |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107909004A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829383A (en) * | 2018-12-29 | 2019-05-31 | 平安科技（深圳）有限公司 | Palm grain identification method, device and computer equipment |
| CN111160306A (en) * | 2019-12-31 | 2020-05-15 | 东南大学 | Three-dimensional palm print recognition method integrating multi-feature and principal component analysis |
| CN111696228A (en) * | 2020-04-22 | 2020-09-22 | 桂林森明智能科技有限责任公司 | Intelligent palm print and palm vein lock based on compressed sensing method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105844261A (en) * | 2016-04-21 | 2016-08-10 | 浙江科技学院 | 3D palmprint sparse representation recognition method based on optimization feature projection matrix |
| CN106446774A (en) * | 2016-08-24 | 2017-02-22 | 施志刚 | Face recognition method based on secondary nearest neighbor sparse reconstruction |
-
2017
- 2017-10-23 CN CN201710996196.8A patent/CN107909004A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105844261A (en) * | 2016-04-21 | 2016-08-10 | 浙江科技学院 | 3D palmprint sparse representation recognition method based on optimization feature projection matrix |
| CN106446774A (en) * | 2016-08-24 | 2017-02-22 | 施志刚 | Face recognition method based on secondary nearest neighbor sparse reconstruction |
Non-Patent Citations (2)
| Title |
|---|
| 曹忠赞: "基于2D和3D掌纹图像方向特征融合的掌纹识别研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 王文龙等: "采用均匀局部二元模式及稀疏表示的掌纹识别", 《光电工程》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829383A (en) * | 2018-12-29 | 2019-05-31 | 平安科技（深圳）有限公司 | Palm grain identification method, device and computer equipment |
| CN109829383B (en) * | 2018-12-29 | 2024-03-15 | 平安科技（深圳）有限公司 | Palmprint recognition method, palmprint recognition device and computer equipment |
| CN111160306A (en) * | 2019-12-31 | 2020-05-15 | 东南大学 | Three-dimensional palm print recognition method integrating multi-feature and principal component analysis |
| CN111160306B (en) * | 2019-12-31 | 2023-11-21 | 东南大学 | Three-dimensional palmprint recognition method integrating multi-feature and principal component analysis |
| CN111696228A (en) * | 2020-04-22 | 2020-09-22 | 桂林森明智能科技有限责任公司 | Intelligent palm print and palm vein lock based on compressed sensing method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ma et al. | Efficient iris recognition by characterizing key local variations | |
| Zaeri | Minutiae-based fingerprint extraction and recognition | |
| CN102332084B (en) | Identity identification method based on palm print and human face feature extraction | |
| CN103279745B (en) | A kind of face identification method based on half face multiple features fusion | |
| Barpanda et al. | Iris feature extraction through wavelet mel-frequency cepstrum coefficients | |
| CN105184254A (en) | Identity authentication method and system | |
| CN102214299A (en) | Method for positioning facial features based on improved ASM (Active Shape Model) algorithm | |
| CN101819629B (en) | Supervising tensor manifold learning-based palmprint identification system and method | |
| CN107909004A (en) | A kind of 3D palmprint recognition technologies | |
| Kekre et al. | Gabor filter based feature vector for dynamic signature recognition | |
| Elnasir et al. | Palm vein recognition based on 2D-discrete wavelet transform and linear discrimination analysis | |
| Zhang et al. | A novel iris recognition method based on feature fusion | |
| Azizi et al. | Efficient iris recognition through improvement of feature extraction and subset selection | |
| Al-Juboori et al. | Biometric authentication system based on palm vein | |
| CN103942545A (en) | Method and device for identifying faces based on bidirectional compressed data space dimension reduction | |
| Premalatha et al. | A dorsal hand vein recognition-based on local gabor phase quantization with whitening transformation | |
| Zhang et al. | Finger vein recognition based on Gabor filter | |
| Guesmi et al. | Curvelet transform-based features extraction for fingerprint identification | |
| Jia et al. | Hand vein recognition algorithm based on NMF with sparsity and clustering property constraints in feature mapping space | |
| Sallehuddin et al. | A survey of iris recognition system | |
| Rafik et al. | A Model Of A Biometric Recognition System Based On The Hough Transform Of Libor Masek and 1-D Log-Gabor Filter | |
| Tang et al. | A review of recent advances in identity identification technology based on biological features | |
| Mukherjee | Indexing techniques for fingerprint and iris databases | |
| Viju et al. | A novel approach to iris recognition for personal authentication | |
| Dehghani et al. | Retinal identification based on rotation invariant moments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: Institute of automation, Heilongjiang Academy of Sciences, 265 Hanshui Road, Nangang District, Harbin City, Heilongjiang Province Applicant after: Institute of intelligent manufacturing, Heilongjiang Academy of Sciences Address before: Institute of automation, Heilongjiang Academy of Sciences, 265 Hanshui Road, Nangang District, Harbin City, Heilongjiang Province Applicant before: INSTITUTE OF AUTOMATION OF HEILONGJIANG ACADEMY OF SCIENCES |
|
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180413 |