CN104268502B - Means of identification after human vein image characteristics extraction - Google Patents
Means of identification after human vein image characteristics extraction Download PDFInfo
- Publication number
- CN104268502B CN104268502B CN201410243187.8A CN201410243187A CN104268502B CN 104268502 B CN104268502 B CN 104268502B CN 201410243187 A CN201410243187 A CN 201410243187A CN 104268502 B CN104268502 B CN 104268502B
- Authority
- CN
- China
- Prior art keywords
- mrow
- vein
- msub
- image
- template
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Security & Cryptography (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses the means of identification after a kind of human vein image characteristics extraction, including the human vein pattern features matrix template being had been stored in using registration process in database, in the process of cognition, first using the vein pattern features matrix of the identified human body of registration process identical method extraction, after obtaining pattern features matrix, the feature of the vein pattern features matrix is matched with the vein pattern features matrix template that each width is stored in database, output matching result, wherein being matched according to equation below： Means of identification after a kind of human vein image characteristics extraction provided by the invention, it is easy and effective, characteristic matching speed is fast, the extracting method of this vein image feature not remove only excessive redundancy, reduce computation complexity and amount of calculation, more effective informations can be included simultaneously, determine that wave filter neighborhood effective range can obtain more preferable recognition effect by building this distance function matrix.
Description
Technical field
The present invention relates to a kind of the identity recognizing technology of human body identity, particularly one are identified by being distinguished to vein pattern
Means of identification after kind human vein image characteristics extraction.
Background technology
In human body identity identification technical field, from single fingerprint recognition to vivo identification, from the rainbow in vivo identification
Membrane technology is to vena metacarpea technology, and all are all in constantly development and perfect.Image characteristics extraction and processing procedure with regard to vena metacarpea
In, also because different extractions and processing means bring different results, so as to directly affects the accuracy of identification and list
One property.As research of the people to human vein feature is found, be not only vena metacarpea, different people its everywhere vein distribution all
There is different changes, how validity feature extraction is carried out to the vein image after extraction, to obtain accurate individual differentiation, and
After extraction, how matching contrast is carried out, will is to be solved in vein identification evolution ask so as to be identified
Topic.
Chinese patent ZL200710144914.5, CN101196987B disclose a kind of " online palmprint, palm vein image
Personal identification method and its special-purpose collection instrument ", vena metacarpea image is normalized first, obtains palm center region
Gray level image, then by the gray level co-occurrence matrixes for the gray level image that palm center region is calculated, by gray level co-occurrence matrixes
The inertial matrix that the gray level image of the central area is calculated is textural characteristics, then calculates the gray level image in palm center region
Each pixel brightness of the luminance mean value as the central area, finally, with reference to palm center area grayscale image
Textural characteristics and brightness, the palm vein image is classified, and then determine whether biopsy sample.The utilization
Palm vein image carries out authentication, is the method classified using the line feature to palm vein blood vessel to realize, head
First the palm center region obtained after normalization is carried out with the multiple dimensioned Gaussian filter of 0 °, 45 °, 90 °, 135 ° four direction
Line detects, and then carries out binaryzation to filtered image, obtains the bianry image of blood vessel line, finally use point to bianry image
The strategy of Point matching is classified, completes authentication.
In file disclosed above, how do not provide by calculating the point-to-point matching obtained to bianry image, and it is high
This wave filter can produce more redundancy or identification information amount deficiency, add computation complexity and the storage of algorithm
Burden, the recognition efficiency and precision of algorithm can be reduced, and vein image is usually present global uneven illumination, lines thickness not
Be unfavorable for feature extraction situations such as, such as and grain boundaries are fuzzy, so how to obtain the venous blood pipeline feature of different live bodies
And how this feature image is handled, to realizing that the authentication of effective different live bodies is particularly important really.
Chinese Patent Application No. be 200710144916.4 " near-infrared image forming apparatus based on palm vein and palmmprint with
Personal identification method ", it is also based on Gauss matched filtering device and convolution is carried out to image, operational formula used by its matching process
Also particularly complicated, identification and time to image, which take, all brings certain influence.
The content of the invention
It is special the invention aims to solve above-mentioned the deficiencies in the prior art and to provide a kind of means of identification easy and effective
Levy the means of identification after the fast human vein image characteristics extraction of matching speed.
To achieve these goals, the means of identification after a kind of human vein image characteristics extraction designed by the present invention,
Including the human vein pattern features matrix template being had been stored in using registration process in database, it is characterized in that recognizing
Cheng Zhong, the vein pattern features matrix of identified human body is extracted using registration process identical method first, obtains directional diagram
After eigenmatrix, palm arteries and veins database is not deposited into, and the feature of the vein pattern features matrix is stored in each width
Vein pattern features matrix template in database is matched, output matching result, wherein will be extracted according to equation below
The vein pattern features matrix template that is stored in database of feature and each width of vein pattern features matrix carry out
Matching：
P represents registration vein template image, and Q represents unknown test image, Pi(x, y) represents what is obtained on i-th of bit
Pattern matrix, Qi(x, y) represents the pattern matrix obtained on i-th of bit, QH, v(x, y) is represented to h, the unknown survey after v translations
Attempt picture, h represents horizontal direction, and v represents vertical direction；M and n represents horizontal and translation pixel count；For XOR, M
× N is image size；Between zero and one, two images are more similar for obtained U values, then closer to 0, conversely, then close to 1；Through
Overmatching, if counted U (P, Q) is less than threshold value set in advance, then the vein pattern extracted matches with the vein template, then
Verification operation success, it is on the contrary then mismatch, authentication failed.
Identification process includes two kinds：First, verify：User to be identified should also provide the identity informations such as ID number, system to system
There to be some palm arteries and veins direction character templates of identical ID number matching in database；2nd, recognize：User to be identified not to
In the case of system provides identity information, direction character template to be measured is matched with all templates of database, system auto-returned
There is the ID number of the training direction character template of maximum matching value with measurement direction feature templates, set in advance if this matching value is more than
Fixed threshold value T, then recognize successfully, otherwise identification failure.
The described vein pattern features matrix that identified human body is extracted with registration process identical method, its method
In, including to human body venous image and establish into coordinate system, vein direction character is then extracted by anisotropic filter,
It is characterized in that：Described is to establishing into coordinate system using Radon conversion by anisotropic filter extraction vein direction character
Image carry out convolution operation, so as to extract vein image feature, feature extraction overall process is as follows：
(1) setpoint distance function：
(i0, j0) centered on pixel, the contiguous range obtained according to the distance function so defined is no more than d picture
The approximate circle region local (i of vegetarian refreshments composition0, j0)。
(2) setting meets the neighborhood of distance function, and divides k direction：
Lk={ (i, j)：J=k (i-i0)+j0, i ∈ Zp(i, j) ∈ local (i0, j0)
Wherein k represents direction, (i0, j0) central pixel point is represented, institute a little (i, j) forms straightway Lk, image is in straight line
Section LkPlace's response is most strong, and other directional responses are 0.
(3) it is corresponding to calculate k oriented energy：
In order to meet it is set meet the neighborhood of distance function, and ensure the picture that the wave filter in k direction is included
Vegetarian refreshments number is identical, and using as far as possible few direction number, ensures direction identification, extract more direction letters as far as possible
Breath, discrimination is improved, when being built in k direction, choose six direction section, wherein：Direction 0 is 0 degree, and the angle of direction 1 belongs to
In [30,31.875] section (((22.5+30)/2+ (30+45)/2)/2=31.875), direction 2 is [58.125,60] (((45
+ 60)/2+ (60+67.5)/2)/2=58.125), direction 3 is 90 degree, and similarly direction 4 is that [120,121.875] are spent in section,
Direction 5 is in [148.125,150] section.
The angle of above-mentioned setting is chosen an angular interval and substitutes accurate angle of the prior art by the present invention, so just
In computer operation, the characteristic information stablized, increase the robustness of feature extraction algorithm.Secondly, angle is substituted with section
Exact value, certain rotation error is may be allowed, strengthen the rotational invariance of algorithm.Finally due to direction 1 and the section ratio of direction 2
30 degree of script and 60 degree of directions can obtain 30 degree of directions closer in 45 degree of directions, the anisotropic filter so set
While upper response, the response on 45 degree of directions is obtained to a certain extent, and similarly the wave filter in direction 3 and direction 4 can obtain
The response in response and a part of 135 degree of directions on 120 degree of directions.Thus reach using as far as possible few direction number,
The purpose of more directional informations is extracted as far as possible, i.e., using the anisotropic filter of same number, can also obtain more more effective
Directional information.
In order to extract more directional informations for more suiting the true framework of vein, the image after convolution operation is carried out to take direction
Index operation, generates response image, i.e. pattern features matrix, and by direction index translation into log2(K) position bit number；Will most
Whole pattern features matrix indexes selection mode as database template, structure intravenous data storehouse, direction described herein, i.e.,
Weighting asks direction index as follows：
P (i, j)=α × k (i, j)+β × o (i, j), α, β ∈ [0,1]
Wherein：Image size is m × n, and k (i, j) is DkK values in (x, y), o (i, j) are the O values in Do (x, y).
The direction index so obtained is exactly corresponding most strong vein direction k and corresponding strong vein direction o weighting
Direction indexes, and under the direction matrix of such formed objects, this direction index selection mode, which can be extracted, more more suits vein
The directional information of true framework.By that analogy, in the case where storage template and speed allow, the direction rope of more intensity responses can be weighted
Draw.
Means of identification after a kind of human vein image characteristics extraction provided by the invention, easy and effective, characteristic matching speed
Degree is fast, and the extracting method of this vein image feature not remove only excessive redundancy, reduce computation complexity and calculating
Amount, while more effective informations can be included, determine wave filter neighborhood effective range energy by building this distance function matrix
Obtain more preferable recognition effect.
Brief description of the drawings
Fig. 1 is the flow chart of vein identification method of the present embodiment based on direction character；
Fig. 2 is ROC curve figure under the different training template numbers of the present embodiment checking test；
Fig. 3 is the FAR and FRR distribution maps under the present embodiment checking test different templates number respectively；
Fig. 4 is the neighborhood schematic diagram for meeting respective distances function for 15 × 15；
Fig. 5 be non-setpoint distance function different k under LkMatrix diagram；
Fig. 6 is L under different k after setpoint distance functionkMatrix diagram；
Fig. 7 is straightway LkFor the matrix diagram under angular interval；
Fig. 8 is 6 anisotropic filter schematic diagrames of the size of embodiment 16 × 16；
Fig. 9 is embodiment direction character template.
Embodiment
The present invention is further described with reference to the accompanying drawings and examples.
Embodiment 1：
As shown in figure 1, the means of identification after a kind of human vein image characteristics extraction that the present embodiment provides, including use
The human vein pattern features matrix template that registration process is had been stored in database, in the process of cognition, first using note
The vein pattern features matrix of the identified human body of volume process identical method extraction, after obtaining pattern features matrix, will not
It, which is stored in, slaps arteries and veins database, and the vein side feature of the vein pattern features matrix and each width being stored in database
Matched to figure eigenmatrix template, output matching result, wherein according to equation below by the vein pattern features of extraction
The feature of matrix is matched with the vein pattern features matrix template that each width is stored in database：
P represents registration vein template image, and Q represents unknown test image, Pi(x, y) represents what is obtained on i-th of bit
Pattern matrix, Qi(x, y) represents the pattern matrix obtained on i-th of bit, QH, v(x, y) is represented to h, the unknown survey after v translations
Attempt picture, h represents horizontal direction, and v represents vertical direction；M and n represents horizontal and translation pixel count；For XOR, M
× N is image size；Between zero and one, two images are more similar for obtained U values, then closer to 0, conversely, then close to 1；Through
Overmatching, if counted U (P, Q) is less than threshold value set in advance, then the vein pattern extracted matches with the vein template, then
Verification operation success, it is on the contrary then mismatch, authentication failed.
Identification process includes two kinds：First, verify：User to be identified should also provide the identity informations such as ID number, system to system
There to be some palm arteries and veins direction character templates of identical ID number matching in database；2nd, recognize：User to be identified not to
In the case of system provides identity information, direction character template to be measured is matched with all templates of database, system auto-returned
There is the ID number of the training direction character template of maximum matching value with measurement direction feature templates, set in advance if this matching value is more than
Fixed threshold value T, then recognize successfully, otherwise identification failure.
As shown in Fig. 2 being ROC curve figure under the different training template numbers of checking test of the present invention, the performance of checking is generally used
Two-dimensional curve is described；Abscissa is FAR, ordinate GAR/FRR.This curve is exactly so-called ROC curve (Receive
Operating Characteristic Curve).ROC curve needs statistical value FAR and FRR, and obtaining the two statistical values needs
Each test image is wanted to be matched with all training samples.If test image comes from same palm with training sample, it
Between matching be referred to as true matching；It is false matching if on the contrary come from different palms.In ROC curve, give one and know by mistake
Rate FAR values, correct receptance GAR value is bigger, illustrates that discrimination is better.From figure 2 it can be seen that when template number is more than or equal to
When 2, for the correct receptance of the present invention generally more than 99.5%, recognition effect is good.And when FAR is big more than or equal to 0.01% template number
When equal to 3, the present invention is correct, and receptance can reach 100%.
As shown in figure 3, it is the FAR and FRR distribution maps, wherein (a) (b) (c) under checking test different templates number respectively
(d) (e) (f) is respectively the FAR and FRR curves under training template is 1,2,3,4,5,6.FAR and FRR has crosspoint in figure,
This point is exactly the point that FAR is equivalent with FRR under some threshold value (equivalent point of this corresponding threshold value is referred to as EER).Traditionally use
EER carrys out the combination property of measure algorithm, for a more excellent palm arteries and veins recognizer, it is desirable under identical threshold condition,
FAR and FRR are the smaller the better.
The running environment of the present embodiment experimental result is Matlab R2010a, Windows XP systems, and CPU is double-core
E5200 (2.5GHz), inside saves as 4GB.Experimental result obtains：
Feature extraction and matching speed ratio compared with
As can be seen from the above table feature extraction speed than [it is slow in NMRT, be into 3 bit institutes by 6 direction encodings
Caused by.But this part be encoded to follow-up matching bring it is convenient, it can be seen that matching speed improves nearly 8 times.And
It is also contemplated that during identification, feature extraction each image need to only be carried out 1 time, and characteristic matching is then with database images number
Increase and increase.From the point of view of the overall time, the time requirement to characteristic matching is more stricter than the time requirement of feature extraction.
Embodiment 2：
Means of identification after a kind of human vein image characteristics extraction that the present embodiment provides, be to provide for it is described with
The vein pattern features matrix of the identified human body of registration process identical method extraction, in its method, including to human vein
Then IMAQ and foundation extract vein direction character into coordinate system by anisotropic filter, described to be filtered by direction
Ripple device extraction vein direction character is to carry out convolution operation to establishing to the image in coordinate system using Radon conversion, so as to carry
Take vein image feature, feature extraction overall process is as follows：
(1) setpoint distance function：
(i0, j0) centered on pixel, the contiguous range obtained according to the distance function so defined is no more than d picture
The approximate circle region local (i of vegetarian refreshments composition0, j0)。
(2) setting meets the neighborhood of distance function, and divides k direction：
Lk={ (i, j)：J=k (i-i0)+j0, i ∈ Zp(i, j) ∈ local (i0, j0)
Wherein k represents direction, (i0, j0) central pixel point is represented, institute a little (i, j) forms straightway Lk, image is in straight line
Section LkPlace's response is most strong, and other directional responses are 0.
(3) it is corresponding to calculate k oriented energy：
Different center pixels repeats this operation.So by D0The direction index matrix that (x, y) is formed is characterized square
Battle array.
Point (i.e. close to the point at edge) generally more remote from center pixel is smaller on center pixel influence, and on the contrary then influence is got over
Greatly, if influenceing size according on center pixel, the pixel for influenceing small is defined as redundancy, the pixel for influenceing big is determined
Justice is effective information.Then the purpose of this setting is to reduce redundancy and as far as possible shadow of the increase effective information to center pixel
Ring.
15 × 15 neighborhood schematic diagram for meeting respective distances function is illustrated in figure 4, black region represents center pixel
Point (i0, j0).Effective range of the most methods using Fig. 1 (a) as its neighborhood, i.e. distance function be d=max (| i-i0|, |
j-j0|), (i0, j0) centered on pixel.The neighborhood that this method defines contains excessive redundancy, adds calculating
Amount and computation complexity.Also effective range of the method using Fig. 1 (b) as its neighborhood, i.e. distance function are d=| i-i0|+|j-
j0|.Although neighborhood amount of calculation and computation complexity that this method defines all accordingly are reduced, its effective information also subtracts significantly
It is few, excessive redundancy is not remove only, while eliminates substantial amounts of effective information.Deficiency based on the above method, herein
Set the new distance function as shown in Fig. 4 (c)：According to the distance so defined
The contiguous range that function obtains is no more than the approximate circle region local (i of d pixel composition0, j0).This method is not
Excessive redundancy is remove only, reduces computation complexity and amount of calculation, while more effective informations can be included.Pass through structure
Build this distance function matrix and determine that wave filter neighborhood effective range can obtain more preferable recognition effect.
Embodiment 3：
Means of identification after a kind of human vein image characteristics extraction that the present embodiment provides, is to meet distance function
Neighborhood, and ensure that the pixel number that the wave filter in k direction is included is identical, and use as far as possible few direction
Number, ensure direction identification, extract more directional informations as far as possible, improve discrimination, when being built in k direction, choose six
Individual Direction interval, wherein：
Direction 0 is 0 degree；
The angle of direction 1 belongs in [30,31.875] section, i.e. ((22.5+30)/2+ (30+45)/2)/2=31.875；
Direction 2 is i.e. ((45+60)/2+ (60+67.5)/2)/2=58.125 in [58.125,60] section；
Direction 3 is 90 degree；
Similarly direction 4 is that [120,121.875] are spent in section；
Direction 5 is in [148.125,150] section.
Such scheme meets following two purposes 1 and purpose 2：
Purpose 1：Each corresponding wave filter in direction, it is assumed that have k anisotropic filter, it should meet following simultaneously
Condition：
Condition 1：It is calculated by formula below：
Lk={ (i, j)：J=k (i-i0)+j0, i ∈ Zp(i, j) ∈ local (i0, j0)
Wherein k represents direction, (i0, j0) central pixel point is represented, institute a little (i, j) forms straightway Lk, image is in straight line
Section LkPlace's response is most strong, and other directional responses are 0.
Condition 2：Ensure that the pixel number that the wave filter in k direction is included is identical.
Meeting the wave filter of condition 1 and condition 2 can obtain from image with straightway LkThe k directional response figures of expression
Picture.
However, how to choose k, when by the wave filter that condition 1 is calculated in direction being not 0 degree and 90 degree, its pixel
Point total number will be total more than 0 degree and 90 degree direction, as shown in figure 5, k=π/6 when form LkShare 42 pixels,
But L is formed during k=0kOnly 32 pixels；In order to meet condition 2, direction must be cast out not on 0 degree and 90 degree
Some point.And the point cast out how is selected, obtained representing the strong corresponding of direction with the point set retained, this point is to discrimination shadow
Sound is larger.But when most methods set wave filter, the direction initialization wave filter of generally artificial sense organ, without strictly in accordance with public affairs
Formula calculates real anisotropic filter, also without the contradiction of discovery condition 12.Distance function above is similarly this part and established
Good basis, due to having cast out redundancy, marginal point of the direction not for 0 degree and 90 degree of wave filter can be cast out, and reduce table
Show the pixel in these directions so that condition 2 is easier to meet, as shown in fig. 6, k=π/6 when form LkShare 33 pixels
Point, L is formed during k=0kHave 32 pixels, now the difference of pixel is only 1.The method of prior art is decile angle, i.e.,
It is spaced equal between any two angle, 0 degree overlaps with 180 degree.
Usual 6 orientation angles are respectively：0 degree, 30 degree, 60 degree, 90 degree, 120 degree, 150 degree；8 orientation angle difference
For：0 degree, 22.5 degree, 45 degree, 67.5 degree, 90 degree, 112.5 degree, 135 degree, 157.5 degree.
Purpose 2：Use as far as possible few more more effective directional informations of anisotropic filter extraction.The methodical filter of institute at present
Ripple device is all to obtain more directional informations by increasing number of filter (direction number).It it is 8 such as by 6 direction increases
Direction, which not only adds amount of calculation, while experimental result is not obviously improved, and has been declined on the contrary.Because phase
With the wave filter of size, the overlapping part in 8 directions adds, and this causes direction can identification decline.This require we
While ensureing direction identification, more directional informations are extracted as far as possible.
The angle of above-mentioned setting is chosen an angular interval and substitutes accurate angle of the prior art by the present embodiment, and this can
With the new rule of point setting to cast out in purpose 1, i.e. straightway LkFor the part overlapped under angular interval, as shown in fig. 7, blue
Part represents that 30 degree of direction represents the non-coincidence part in 30 degree of directions, green portion with 31.875 degree of direction intersection, yl moiety
Divide and represent 31.875 degree of non-coincidence parts, crocus is whole neighborhood region, and black portions represent central pixel point, so by indigo plant
It is just 32 pixels that color and black portions, which combine the pixel total number to be formed, and this is total with pixel on 0 degree of 90 degree of direction
Number is identical, is so easy to computer operation, the characteristic information stablized, and increases the robustness of feature extraction algorithm.Secondly, angle
Degree substitutes exact value with section, may be allowed certain rotation error, strengthens the rotational invariance of algorithm.Finally due to direction 1
With the section of direction 2 than 30 degree of script and 60 degree of directions are closer in 45 degree of directions, the anisotropic filter so set can be
While obtaining responding on 30 degree of directions, the response on 45 degree of directions is obtained to a certain extent, similarly the filter in direction 3 and direction 4
Ripple device can obtain the response in response and a part of 135 degree of directions on 120 degree of directions.Thus reach purpose 2, even if
With the anisotropic filter of same number, more more effective directional informations are obtained.
As shown in figure 8, it is 6 anisotropic filter schematic diagrames of the size of the present embodiment 16 × 16, in Fig. 8 anisotropic filter
In, line width 2, direction quantity is 6, and (a) (b) (c) (d) (e) (f) image represents the line integral (summation) of different directions respectively,
These directions are respectively 0 °, [30 °, 31.875 °], [58.125 °, 60 °], 90 °, [120 °, 121.875 °], [148.125 °,
150°].Using this anisotropic filter, the direction character of 4 pixels (center pixel) can be calculated every time.
Embodiment 4：
Means of identification after a kind of human vein image characteristics extraction that the present embodiment provides, is more more cut to extract
The directional information of the true framework of vein is closed, the image after convolution operation is carried out to take direction index operation, generates response image, i.e.,
Pattern features matrix, and by direction index translation into log2(K) position bit number；Using final pattern features matrix as number
According to library template, structure intravenous data storehouse.
Pattern features matrix is expressed as：
In the vein image collected, represent that the gray value of venosomes is darker than the gray value at background place, therefore Energy
[Lk] smaller, represent that this direction vein is corresponding most strong.But direction index operation is taken there are a variety of methods according to vein response, at present
Most methods choose the corresponding most strong direction of vein, i.e. Energy [Lk] minimum direction index construct pattern features square
Battle array：
Wherein image size is m × n, and k (i, j) is DkK values in (x, y).But the letter that this pattern matrix is included
It is the most strong direction of vein response to cease, in characteristic extraction procedure, LkAdjustable (i.e. center pixel (the i of line width0, j0) can be not only
Only 1), in but once testing, LkLine width value is fixed, it is impossible to is changed halfway with vein change width.If vein width
Change greatly, or vein crossings point, then individually most strong direction index can not completely express vein directional information or show
False vein directional information.
Therefore, a kind of direction index selection mode that the present embodiment provides, i.e. weighting ask direction index as follows：
P (i, j)=α × k (i, j)+β × o (i, j), α, β ∈ [0,1]
Wherein：Image size is m × n, and k (i, j) is DkK values in (x, y), o (i, j) they are the o values in Do (x, y), this
The direction index that sample obtains is exactly that corresponding most strong vein direction k and corresponding strong vein direction o weighting direction index, this
Under the direction matrix of sample formed objects, this direction index selection mode can extract more sides for more suiting the true framework of vein
To information.By that analogy, in the case where storage template and speed allow, the direction index of more intensity responses can be weighted.
In the present embodiment, wave filter size is 28~36；Direction number is set to 6, and line width is 2~7.Indexed according to different directions
Selection Strategy represents vein direction character, forms vein direction character template as shown in Figure 9, can from direction character template
Venous structures feature can preferably be reflected by going out the present embodiment feature extraction mode, wherein each lattice represents a pixel, no
Different direction value is represented with gray value.
Claims (1)
1. the means of identification after a kind of human vein image characteristics extraction, including had been stored in using registration process in database
Human vein pattern features matrix template, it is characterized in that in the process of cognition, carried first using registration process identical method
The vein pattern features matrix of identified human body is taken, after obtaining pattern features matrix, by the vein pattern features matrix
Feature matched with the vein pattern features matrix template that each width is stored in database, output matching result, its
The middle vein side being stored in the feature of the vein pattern features matrix of extraction and each width according to equation below in database
Matched to figure eigenmatrix template：
<mrow>
<mi>U</mi>
<mrow>
<mo>(</mo>
<mi>P</mi>
<mo>,</mo>
<mi>Q</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mrow>
<mo>&ForAll;</mo>
<mi>h</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mo>-</mo>
<mi>m</mi>
<mo>,</mo>
<mi>m</mi>
<mo>&rsqb;</mo>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>v</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mo>-</mo>
<mi>n</mi>
<mo>,</mo>
<mi>n</mi>
<mo>&rsqb;</mo>
</mrow>
</munder>
<mo>{</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mi>M</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mi>N</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mn>3</mn>
</munderover>
<mrow>
<mo>(</mo>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
</mrow>
<mo>)</mo>
<mo>&CircleTimes;</mo>
<msubsup>
<mi>Q</mi>
<mi>i</mi>
<mrow>
<mi>h</mi>
<mo>,</mo>
<mi>v</mi>
</mrow>
</msubsup>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mn>3</mn>
<mo>&times;</mo>
<mi>M</mi>
<mo>&times;</mo>
<mi>N</mi>
</mrow>
</mfrac>
<mo>}</mo>
</mrow>
P represents registration vein template image, and Q represents unknown test image, Pi(x, y) represents the template square obtained on i-th of bit
Battle array, Qi(x, y) represents the pattern matrix obtained on i-th of bit, Qh,v(x, y) is represented to h, the unknown test chart after v translations
Picture, h represent horizontal direction, and v represents vertical direction；M and n represents horizontal and translation pixel count；For XOR, M × N is
Image size；Between zero and one, two images are more similar for obtained U values, then closer to 0, conversely, then close to 1；By
Match somebody with somebody, if counted U (P, Q) is less than threshold value set in advance, then the vein pattern extracted matches with the vein template, then verified
Operate successfully, it is on the contrary then mismatch, authentication failed；
Wherein identification process includes two kinds：First, verify：User to be identified should also provide the identity informations such as ID number, system to system
There to be some palm arteries and veins direction character templates of identical ID number matching in database；2nd, recognize：User to be identified not to
In the case of system provides identity information, direction character template to be measured is matched with all templates of database, system auto-returned
There is the ID number of the training direction character template of maximum matching value with measurement direction feature templates, set in advance if this matching value is more than
Fixed threshold value T, then recognize successfully, otherwise identification failure；
The described vein pattern features matrix that identified human body is extracted using registration process identical method, including to human body
Then vein image acquisition and foundation pass through anisotropic filter and extract vein direction character, the described side of passing through into coordinate system
It is that convolution operation is carried out to establishing to the image in coordinate system using Radon conversion to wave filter extraction vein direction character, from
And vein image feature is extracted, feature extraction overall process is as follows：
(1) setpoint distance function：
<mrow>
<mi>d</mi>
<mo>=</mo>
<mi>r</mi>
<mi>o</mi>
<mi>u</mi>
<mi>n</mi>
<mi>d</mi>
<mrow>
<mo>(</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>-</mo>
<msub>
<mi>i</mi>
<mn>0</mn>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>-</mo>
<msub>
<mi>j</mi>
<mn>0</mn>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>)</mo>
</mrow>
</mrow>
(i0,j0) centered on pixel, the contiguous range obtained according to the distance function so defined is no more than d pixel
The approximate circle region local (i of composition0,j0)；
(2) setting meets the neighborhood of distance function, and divides k direction：
Lk={ (i, j):J=k (i-i0)+j0,i∈Zp}(i,j)∈local(i0,j0)
Wherein k represents direction, (i0,j0) central pixel point is represented, institute a little (i, j) forms straightway Lk, image is in straightway Lk
Place's response is most strong, and other directional responses are 0；
(3) k oriented energy response is calculated：
<mrow>
<mi>E</mi>
<mi>n</mi>
<mi>e</mi>
<mi>r</mi>
<mi>g</mi>
<mi>y</mi>
<mo>&lsqb;</mo>
<msub>
<mi>L</mi>
<mi>k</mi>
</msub>
<mo>&rsqb;</mo>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
<mo>&Element;</mo>
<msub>
<mi>L</mi>
<mi>k</mi>
</msub>
</mrow>
</munder>
<mi>f</mi>
<mo>&lsqb;</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>&rsqb;</mo>
<mo>;</mo>
</mrow>
When being built in k direction, six direction section is chosen, wherein：
Direction 0 is 0 degree；
The angle of direction 1 belongs in [30,31.875] section；
Direction 2 is [58.125,60]；
Direction 3 is 90 degree；
Direction 4 is that [120,121.875] are spent in section；
Direction 5 is in [148.125,150] section；
It is described to calculate in k oriented energy response, direction index selection mode, ask direction to index using weighting：
<mrow>
<msub>
<mi>D</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>arg</mi>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mi>k</mi>
</munder>
<mrow>
<mo>(</mo>
<mi>E</mi>
<mi>n</mi>
<mi>e</mi>
<mi>r</mi>
<mi>g</mi>
<mi>y</mi>
<mo>&lsqb;</mo>
<msub>
<mi>L</mi>
<mi>k</mi>
</msub>
<mo>&rsqb;</mo>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>K</mi>
</mrow>
<mrow>
<msub>
<mi>D</mi>
<mi>o</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>arg</mi>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mi>o</mi>
</munder>
<mrow>
<mo>(</mo>
<mi>E</mi>
<mi>n</mi>
<mi>e</mi>
<mi>r</mi>
<mi>g</mi>
<mi>y</mi>
<mo>&lsqb;</mo>
<msub>
<mi>L</mi>
<mi>o</mi>
</msub>
<mo>&rsqb;</mo>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>o</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>K</mi>
<mo>,</mo>
<mi>o</mi>
<mo>&NotEqual;</mo>
<mi>k</mi>
</mrow>
P (i, j)=α × k (i, j)+β × o (i, j), α, β ∈ [0,1]
Wherein：Image size is m × n, and k (i, j) is DkK values in (x, y), o (i, j) are the o values in Do (x, y).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410243187.8A CN104268502B (en) | 2013-06-02 | 2014-06-03 | Means of identification after human vein image characteristics extraction |
Applications Claiming Priority (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013102138721 | 2013-06-02 | ||
| CN201310213872 | 2013-06-02 | ||
| CN201310213872.1 | 2013-06-02 | ||
| CN201410243187.8A CN104268502B (en) | 2013-06-02 | 2014-06-03 | Means of identification after human vein image characteristics extraction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104268502A CN104268502A (en) | 2015-01-07 |
| CN104268502B true CN104268502B (en) | 2018-01-19 |
Family
ID=52160022
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410243187.8A Active CN104268502B (en) | 2013-06-02 | 2014-06-03 | Means of identification after human vein image characteristics extraction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104268502B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975960B (en) * | 2016-06-16 | 2019-03-15 | 湖北润宏科技股份有限公司 | Iris identification method based on grain direction energy feature |
| CN108509886B (en) * | 2018-03-26 | 2021-08-17 | 电子科技大学 | Palm vein identification method based on vein pixel point judgment |
| CN112288660B (en) * | 2020-12-25 | 2021-04-13 | 四川圣点世纪科技有限公司 | Vein image restoration method and device based on NMRT (NMRT) directional feature constraint |
| CN114241534B (en) * | 2021-12-01 | 2022-10-18 | 佛山市红狐物联网科技有限公司 | Rapid matching method and system for full-palm venation data |
| CN118116036A (en) * | 2024-01-25 | 2024-05-31 | 重庆工商大学 | Finger vein image feature extraction and coding method based on deep reinforcement learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101055618A (en) * | 2007-06-21 | 2007-10-17 | 中国科学院合肥物质科学研究院 | Palm grain identification method based on direction character |
| EP2137673A1 (en) * | 2007-05-31 | 2009-12-30 | Aisin AW Co., Ltd. | Feature extraction method, and image recognition method and feature database creation method using the same |
-
2014
- 2014-06-03 CN CN201410243187.8A patent/CN104268502B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2137673A1 (en) * | 2007-05-31 | 2009-12-30 | Aisin AW Co., Ltd. | Feature extraction method, and image recognition method and feature database creation method using the same |
| CN101055618A (en) * | 2007-06-21 | 2007-10-17 | 中国科学院合肥物质科学研究院 | Palm grain identification method based on direction character |
Non-Patent Citations (1)
| Title |
|---|
| On the injectivity of the circular Radon transform arising in thermoacoustic tomography;Gaik Ambartsoumian etc;《Inverse Problems》;20050228;第21卷(第2期);摘要、第4页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104268502A (en) | 2015-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bhanu et al. | Deep learning for biometrics | |
| CN104268502B (en) | Means of identification after human vein image characteristics extraction | |
| CN105760841B (en) | Identity recognition method and system | |
| CN101093538B (en) | Method for identifying iris based on zero crossing indication of wavelet transforms | |
| CN110427832A (en) | A kind of small data set finger vein identification method neural network based | |
| Feng et al. | Robust and efficient algorithms for separating latent overlapped fingerprints | |
| US20060008124A1 (en) | Iris image-based recognition system | |
| WO2017059591A1 (en) | Finger vein identification method and device | |
| CN101281600B (en) | Method for acquiring palm print characteristics as well as corresponding personal identification method based on palm print | |
| CN102073843A (en) | Non-contact rapid hand multimodal information fusion identification method | |
| CN104091145B (en) | Human body slaps arteries and veins characteristic image acquisition method | |
| CN107066961B (en) | Fingerprint method for registering and device | |
| CN107563328A (en) | A kind of face identification method and system based under complex environment | |
| CN107066934A (en) | Tumor stomach cell image recognition decision maker, method and tumor stomach section identification decision equipment | |
| CN111462379A (en) | Access control management method, system and medium containing palm vein and face recognition | |
| CN107392082A (en) | A kind of small area fingerprint comparison method based on deep learning | |
| Wyzykowski et al. | Level three synthetic fingerprint generation | |
| TW201211913A (en) | A method for recognizing the identity of user by palm vein biometric | |
| CN103218624A (en) | Recognition method and recognition device based on biological characteristics | |
| CN109583279A (en) | A kind of fingerprint and refer to that vein combines recognizer | |
| Liu et al. | Fingerprint presentation attack detector using global-local model | |
| CN103324921B (en) | A kind of mobile identification method based on interior finger band and mobile identification equipment thereof | |
| Yuan et al. | Fingerprint liveness detection using histogram of oriented gradient based texture feature | |
| Lefkovits et al. | CNN approaches for dorsal hand vein based identification | |
| Al-Juboori et al. | Biometric authentication system based on palm vein |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 516008 305, third Tsing Hua court, 11 North lace Road, Huizhou, Guangdong. Patentee after: GUANGDONG WICROWN INFORMATION TECHNOLOGY Co.,Ltd. Address before: 516000, 310 floor, Tsinghua court third, 11 Hua Bian North Road, Huizhou, Guangdong. Patentee before: GUANGDONG WICROWN INDUSTRIAL DEVELOPMENT Co.,Ltd. |
|
| CP03 | Change of name, title or address | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220307 Address after: 518109 1011, Taojing Jinhua building, Taoyuan community, Dalang street, Longhua District, Shenzhen, Guangdong Province Patentee after: Zhangmaitong Information Technology Co.,Ltd. Address before: 516008 305, third Tsing Hua court, 11 North lace Road, Huizhou, Guangdong. Patentee before: GUANGDONG WICROWN INFORMATION TECHNOLOGY Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230407 Address after: 518042 303, Floor 3, Building 6, Anhua Community, No. 18, Tairan 8th Road, Tian'an Community, Shatou Street, Futian District, Shenzhen, Guangdong Patentee after: Zhiguan Yizhangtong Technology (Shenzhen) Co.,Ltd. Address before: 518109 1011, Taojing Jinhua building, Taoyuan community, Dalang street, Longhua District, Shenzhen, Guangdong Province Patentee before: Zhangmaitong Information Technology Co.,Ltd. |
|
| TR01 | Transfer of patent right |