CN1327387C - Method for identifying multi-characteristic of fingerprint - Google Patents

Abstract

The present invention relates to a fingerprint multi-characteristic recognition method which belongs to the technical field of fingerprint recognition. The present invention is characterized in that the present invention combines traditional detail point characteristics and directional figure model parameters together and uses the traditional detailed point characteristics and directional figure model parameters as fingerprint recognition characteristics; in addition, a directional comparison result and a traditional detail point comparison result are combined so as to greatly improve fingerprint recognition rate. Besides, the method saves storage space and is convenient and practical. The present invention is particularly suitable for adults and people with bad fingerprint quality.

Description

The many characteristic recognition methods of fingerprint

Technical field

The present invention relates to the fingerprint identification technology field, relate in particular to and merge the technology that directional diagram feature and minutiae feature are discerned simultaneously.

Background technology

Development and progress along with society, carry out fast, effectively, the actual requirement distinguished of the person automatically is urgent day by day, the turnover of important department security personnel, the control of passing by, immigration inspection, secret or valuables place of retention, prevents that credit card deception, network security etc. from all needing to carry out personal reliably discriminating.In the foundation of authentication, key, certificate may lose, stolen or duplicate, password is forgotten about, is obscured or stolen a glance at easily again, and biological characteristic (comprising fingerprint, people's face, hand shape, handwritten signature, iris etc.) is people's a inherent attribute, above-mentioned situation can not occur, therefore become the optimal foundation of distinguishing.Wherein, fingerprint recognition is to use the most general, the highest, the easiest received personal identification identification of discrimination.First of material evidence, the existing history very very long and achievement is abundant of fingerprint recognition.The automatic fingerprint recognition of computer based starts from twentieth century sixties, and it at first is applied in the criminal detection.In recent years, fingerprint automation recognition is generalized to fields such as work attendance, gate inhibition, safe deposit box, social insurance gradually from criminal application, and China determines tentatively that also introducing finger print information in the new breed of identity cards carries out personal identification.After 911 terrorist incidents, more obtain unprecedented attention based on the person identification of fingerprint automation recognition.

Present fingerprint automation recognition method mainly is based on minutiae feature, promptly extracts minutiae point (destination node of crestal line or point of crossing in the fingerprint) and characterizes fingerprint image as feature, discerns by comparing these features.Its step generally comprises: fingerprint image acquisition, directional diagram (direction of fingerprint texture) extraction, figure image intensifying, fingerprint ridge line refinement, minutiae point extraction etc.Round how extracting minutiae point faster and better, recent two decades comes domestic and international research unit to make extensive work, and existing fingerprint product all is based on this method, SecuTouch as U.S. BAC company, the FIU-500 of Japan Sony company, the Veriprox of U.S. BII company, the Bogo2000 of Korea S Bogo company, the U.are.U 2000 of U.S. DP company, the Biologon of American I dentix company etc.

Along with the popularization that fingerprint recognition is used, existing system and the method deficiency on widespread adoption more and more displays.The one piece of paper (S.Pankanti that delivers recently, S.Prabhakar, and A.K.Jain.On the individuality offingerprints.Proceedings of IEEE International Symposium of CVPR, Vol.1, USA, Dec2001, pp:805-812) point out, on average can only obtain 36 genuine minutiae point for one piece of fingerprint in present a lot of system of fingerprints, and it has been generally acknowledged that, just can be judged as same finger if there is minutiae point more than 15 to compare successfully between two pieces of fingerprints, consider that the minutiae point that extracts with computer approach certain error can occur on coordinate and direction, the probability that then fingerprint of some fingers is judged as another is greater than 4.26 * 10 ^-7(having provided the example that two pieces of different fingerprints are judged as same finger among Fig. 1).If considering has false point (all this situation will inevitably occur for any method) in the minutiae point that extracts, then error rate also can increase greatly, reaches 10 ^-3Degree, have a strong impact on the result of identification.

A basic reason that above problem occurs is that the feature representation of fingerprint is complete inadequately.The single description with minutiae point of the pattern of this more complicated of fingerprint is far from being enough.We can say that this problem has become a bottleneck problem of current fingerprint recognition.Address this problem, must restudy the feature representation problem of fingerprint, seek the new feature that can describe fingerprint individual difference and self stability from brand-new angle.Certainly, because the restriction of practical application, these new features should have brief expression (with convenient storage).In all patents that can find or deliver in the document, with more approaching the having of thinking of the present invention: at (A.K.Jain, S.Prabhakar, L.Hong and S.Pankanti.Filterbank-based fingerprint matching.IEEE Trans.on Image Processing, Vol.9,2000, pp:846-859) utilize the response of multi-filter to encode in, carry out fingerprint recognition with these codings as feature, at (A.Ross, S.Prabhakar, and A.K.Jain.Fingerprint matching using minutiae and texture features.Proc.ICIP 2001, Vol.3, Greece, Oct 2001, pp 282-285) adopt in above-mentioned thinking similarly coding add that minutiae point information carries out fingerprint recognition, all obtained certain effect, but its problem is to need storage space very big, causes the obstacle of application.Improvement effect neither be very desirable.The inventive method is then utilized the feature of directional diagram as new fingerprint representation, with the directional diagram model tormulation, only needs the storage few parameters; When fingerprint recognition minutiae point and directional diagram model parameter are together used, the storage space of required increase is far smaller than preceding method, and feature is more reliable and more stable.And, preserve the technical scheme that the directional diagram parameter helps designing new practicality because directional diagram all will be used in a plurality of steps of fingerprint recognition.

Summary of the invention

The fingerprint representation method and the corresponding recognition technology that the purpose of this invention is to provide the practicality of many features are used in the time of can preserving more Useful Informations for fingerprint comparison when the storage fingerprint characteristic.The present invention adopts model that fingerprint orientation is described, and calculates the directional diagram model parameter of any fingerprint; In the fingerprint comparison stage, the model parameter of the enough storages of energy is recovered fingerprint orientation, and then carries out the directional diagram comparison of any two fingerprints, combines with the minutiae point comparison result, thereby reaches the purpose that improves the traditional recognition method effect.

Core concept of the present invention is that very important directional diagram in the fingerprint recognition is only expressed with few parameters, so just can store these parameters and use in the comparison stage, just combines with traditional minutiae point recognition methods and can improve recognition performance.In fingerprint, what minutiae point reflected is the minutia of fingerprint, and what directional diagram reflected is global characteristics, has very big complementarity.Because directional diagram all is continuous except that several points (singular point), therefore can adopt the weighted sum of binary polynomial and singular function to express directional diagram again.Adopt least square method can obtain the parameter of above-mentioned model.Adopt traditional fusion method can realize the combination of minutiae feature and directional diagram feature.Wherein based on minutiae feature recognition methods can select any method of present use for use.The present invention is only to a kind of introduction the wherein.

The fingerprint representation feature that we propose comprises: singular point (coordinate), minutiae point (coordinate, direction), directional diagram parameter, effective coverage chain code.Classic method is mainly used preceding two kinds, and we then increase and have used the directional diagram parameter, the effective coverage chain code.These features and original feature have big independence, merge to get up can improve discrimination greatly.

Above-mentioned feature can be stored with the block code mode.Below introduced respectively:

Singular point: i.e. directional diagram discrete point is divided into central point and trigpoint.General central point is less than 3, and trigpoint is less than 3.Store the x coordinate of these singular points, the y coordinate, each coordinate is deposited (reason is seen the minutiae point storage area) with a byte.So need 12 bytes altogether.

Minutiae point: comprise the x coordinate of minutiae point, y coordinate and orientation angle.For saving storage space, half (with fingerprint image size (512 * 320) is example, and the minutiae point of storage and the coordinate range of singular point are 0-255, and the scope of direction value is between 0 to 180) of storage actual value size.Therefore, minutiae point need be with 3 byte representations.Generally speaking, the minutiae point quantity of fingerprint is below 100, so need 300 byte representations at most;

Directional diagram: the model parameter of directional diagram.With the mixture model is example, and it is 4 that the multinomial model part is generally got n, promptly uses two four binary polynomial matches, and each multinomial coefficient is 25 parameters, so polynomial segment has 50 of parameters, with float type (4 byte) storage, needs 200 bytes; The point charge model part only needs storing electricity, also with the storage of float type, because of singular point is less than 6, needs 24 bytes store altogether.

The effective coverage: need storage border, fingerprint effective coverage, adopt ten hexagon method for expressing, storage summit x coordinate, the y coordinate, similar with minutiae point, half of storage actual coordinate value needs 32 bytes altogether.

To sum up, need 568 bytes store fingerprint characteristics altogether.If do not store minutiae point direction value (can release), then need 468 bytes altogether from the directional diagram feature.

Describe algorithm for recognizing fingerprint below in detail based on many features.The key step of feature extraction phases comprises: extract the effective coverage, and the field of direction estimates that singular point detects, directional diagram modeling Flame Image Process and enhancing, and minutiae point is extracted; Feature relatively is that the key step of cognitive phase comprises: minutiae point comparison, directional diagram comparison, Decision Fusion.Wherein, effective coverage extraction, field of direction estimation, Flame Image Process all can use conventional methods than each step of equity with enhancing, minutiae point extraction and minutiae point carries out.Below each step is made introductions all round.

Feature extraction

Extract the effective coverage

Refer to that by getting fingerprint partly is not to be full of full figure among the original fingerprint figure that device collects, the parts of images that contains fingerprint is just meaningful in fingerprint recognition, is called the effective coverage.Original image is divided into size is the grid of (4 * 4),, calculate the average and the variance of the gray-scale value of all pixels in this zone, think that just this point is in the effective coverage when having only both all to satisfy separately condition each such grid.Formula below wherein the calculating of average and variance relies on:

$I_{\mathrm{avg}}(i,j)=\frac{1}{16}\underset{x=0}{\overset{3}{\mathrm{\Σ}}}\underset{y=0}{\overset{3}{\mathrm{\Σ}}}I(i+x,j+y);$

$\mathrm{Var}(i,j)=\frac{1}{16}\underset{x=0}{\overset{3}{\mathrm{\Σ}}}\underset{y=0}{\overset{3}{\mathrm{\Σ}}}{(I(i+x,j+y)-I_{\mathrm{avg}}(i,j))}^2;$

Here, I _Avg(i, j), (i j) is illustrated respectively in that (i j) is the gray average and the variance of the grid in the upper left corner, and (i+x j+y) is (i+x, gradation of image value j+y) to I with point to Var.Require as th1＜I _Avg(i, j)＜th2 and Var (i, j)＞during th3, this grid of mark is effectively, is labeled as 1.Wherein, threshold value is chosen as: th1=20; Th2=220; Th3=6.

Grids all on the image is carried out aforesaid operations,, need carry out aftertreatment in order to remove noise effect:

1,3 * 3 filtering, specific practice are exactly to check to comprise 9 points of tested measuring point in its interior 3 * 3 neighborhood, are that effectively other all are invalid, think that so this point is a noise if having only this point, and change is labeled as 0 (it is invalid to show); If it is invalid having only this point, other all are effectively, think that so this point is an available point, and change was labeled as for 1 (showing effectively).

2, remove " hole " in the middle of the effective coverage, method is to line by line scan, and fills up all Null Spots between the Far Left and rightmost available point in the image, and it is labeled as effective 1; By column scan, fill up in the image topmost and all Null Spots between the available point bottom, it is labeled as effective 1.

So just obtained effective coverage (length and width be respectively former figure 1/4).As Figure 11-2).

In the comparison based on the field of direction, the effective coverage plays an important role, and therefore need preserve.For saving storage space, we adopt one ten hexagon to come approximate representation fingerprint image effective coverage.Concrete operations are as follows:

Obtain the barycenter (x in authentic and valid zone ₀, y ₀):

$x_0=\frac{1}{N}\underset{(i,j)\&Element;E}{\mathrm{\&Sigma;}}i,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">y</figure-callout>}}_0=\frac{1}{N}\underset{(i,j)\&Element;E}{\mathrm{\&Sigma;}}j$

Wherein, E represents the effective coverage, and N represents counting in the effective coverage

From barycenter (x ₀, y ₀) locate equal angular to around 16 directions (be respectively 0, π/8, π/4 ..., 15 π/8) and the injection line, obtain the intersection point of these rays and effective coverage outer boundary.In the time of specific implementation,, illustrate that then this point is the border intersection point as long as check that the point on the ray is invalid by effectively becoming.We only need the coordinate of these intersection points of storage, and promptly ten hexagonal summits just can recover approximate effective coverage.

By Figure 11-3) as can be seen, Biao Shi effective coverage and authentic and valid regional basically identical in this way.Experimental result shows that also both difference are minimum to the comparison influence of back.

The field of direction is estimated

The field of direction is the piece image of expression fingerprint ridge line trend, and wherein every numerical value has been represented the local crestal line direction of corresponding point in the fingerprint image.Directional diagram has been portrayed the global information of fingerprint, plays an important role in fingerprint recognition.The algorithm that is based on the gradient statistics that adopts in this method, effect is as Figure 11-7) shown in.Algorithm is as follows:

1, utilizes the horizontal direction operator S of Soble operator _xWith vertical direction operator S _y(see figure 4) ask for point (x, shade of gray y):

Horizontal direction: $G_x(x,y)=\frac{1}{8}{\mathrm{\&Sigma;}}_{j=y-1}^{y+1}{\mathrm{\&Sigma;}}_{i=x-1}^{x+1}{S}_x(x-i,y-j)I(i,j);$

Vertical direction: $G_y(x,y)=\frac{1}{8}{\mathrm{\&Sigma;}}_{j=y-1}^{y+1}{\mathrm{\&Sigma;}}_{i=x-1}^{x+1}{S}_y(x-i,y-j)I(i,j);$

Wherein (i j) is (i, gray-scale value j) to I.

2, fingerprint image being divided into size is Grid, $\stackrel{\&OverBar;}{W}=7,$ Ask for the local direction θ of each grid correspondence with following formula:

$\mathrm{\&theta;}(i,j)=\frac{1}{2}{\tan }^{-1}\left(\frac{{\mathrm{\&Sigma;}}_{i=1}^{\stackrel{\&OverBar;}{W}}{\mathrm{\&Sigma;}}_{j=1}^{\stackrel{\&OverBar;}{W}}{2G}_x(i,j)G_y(i,j)}{{\mathrm{\&Sigma;}}_{i=1}^{\stackrel{\&OverBar;}{W}}{\mathrm{\&Sigma;}}_{j=1}^{\stackrel{\&OverBar;}{W}}(G_x^2(i,j)-G_y^2(i,j))}\right);$

Certainty level of directional field

Certainty level of directional field is the further result that the field of direction is estimated.For each point in the field of direction, calculate its consistance of the interior field of direction of small neighbourhood on every side, and, be referred to as certainty level of directional field its degree of reiability as this point.Its value is between [0,1].Certainty level of directional field has reflected topography's quality to a certain extent, the zone that picture quality is good more, and certainty level of directional field is high more, the zone that picture quality is poor more, certainty level of directional field is low more.Effect is as Figure 11-8) shown in.Acquiring method is as follows:

$c(i,j)=\frac{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x^2(i+x,j+y)-G_y^2(i+x,j+y))\right)}^2}{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x^2(i+x,j+y)+G_y^2(i+x,j+y))\right)}^2}$

$+\frac{4\&times;{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x(i+x,j+y)\&times;G_y(i+x,j+y))\right)}^2}{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x^2(i+x,j+y)+G_y^2(i+x,j+y))\right)}^2};$

Wherein c (i, j) promptly point (i, certainty level of directional field j), $\stackrel{\&OverBar;}{W}=7.$

To use as weighting coefficient in the directional diagram modeling operation that certainty level of directional field will be mentioned below, concrete grammar has detailed description in this step.

Singular point detects

Utilize the Poincare index method on the field of direction, to carry out singular point and detect, according to principle be: the Poincare index of general point is 0, and the Poincare index of central point is 1/2, and the Poincare index of trigpoint is-1/2.Point (i, j) computation process of the Poincare index on is as follows:

Ψ _x(k) and Ψ _y(k) represent with the set point to be that radius centered is 8 circle (total N on the circle respectively _ψHorizontal stroke, the ordinate of k point=25 picture elements).Direction adjacent 2 on the circle is changed to:

Wherein: δ (k)=O (Ψ _x(k '), Ψ _y(k '))-O (Ψ _x(k), Ψ _y(k));

k′＝(k+1)modN _ψ；

Here the field of direction of O () expression fingerprint image (codomain be [0,180)).

Point (i, j) the Poincare index on is:

$\mathrm{Poincare}(i,j)=\frac{1}{360}\underset{k=0}{\overset{N_{\mathrm{\&psi;}}-1}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}\left(k\right);$

Field of direction modeling

The direction of fingerprint field reflects the information of overall importance of fingerprint, and is extremely important to fingerprint recognition.The common method that is based on gradient that adopts, it has two disadvantages: the field of direction of 1, obtaining is affected by noise bigger; 2, can not direct memory access figure, reason is that requisite space is excessive.Therefore need carry out modeling to the direction of fingerprint field, can make the directional diagram of trying to achieve more reliable like this, and the storage space that needs also can be very little.The mixture model that we adopt comprises two parts, and a part is a multinomial model, promptly to the field of direction of whole fingerprint, portrays with two binary polynomials of degree n; Another part is a point charge model, promptly near singular point not in the field of direction of smooth domain, revise with the standard charge model, model is as follows:

$\left(\begin{array}{c}R(x,y)\\ I(x,y)\end{array}\right){=\mathrm{\&alpha;}}_{\mathrm{PM}}\&CenterDot;\left(\begin{array}{c}\mathrm{PR}(x,y)\\ \mathrm{PI}(x,y)\end{array}\right)+\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{PC}}^{\left(k\right)}\&CenterDot;\left(\begin{array}{c}{H}_1^{\left(k\right)}(x,y)\\ H_2^{\left(k\right)}(x,y)\end{array}\right);$

Wherein R (x, y), I (x, y) be mixture model point (x, value y), PR (x, y), (x is that multinomial model is at point (x, value y), H y) to PI ₁ ^(k)(x, y), H ₂ ^(k)(x is that the point charge model of k singular point correspondence is in point (x, value y) y); α _PC ^(k)(x, y) expression k singular point correspondence point charge model point (computing formula is for x, the weight of y) locating: ${\mathrm{\&alpha;}}_{\mathrm{Pc}}^{\left(k\right)}(x,y)=1-\frac{r^{\left(k\right)}(x,y)}{R^K},$ $r^{\left(k\right)}(x,y)=\min (\sqrt{{(x-x_0^{\left(k\right)})}^2+{(y-y_0^{\left(k\right)})}^2},R^{\left(k\right)}),$ R ^(k)Be the effective radius of the point charge model of k singular point correspondence, (x ₀ ^(k), y ₀ ^(k)) be the coordinate of k singular point; The weight of multinomial model in mixture model is ${\mathrm{\&alpha;}}_{\mathrm{PM}}(x,y)=\max \{1-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{Pc}}^{\left(k\right)}(x,y),0\}.$

Near point central point (x, y) the standard point charge model on is:

${\mathrm{PC}}_{\mathrm{Core}}=H_1+i{H}_2=\left\{\begin{array}{cc}\frac{y-y_0}{r}Q-i\frac{x-x_0}{r}Q,& r\&le;R\\ 0,& r>R\end{array}\right.;$

(x wherein ₀, y ₀) be center point coordinate, Q is the point charge electric weight, and R is that point charge influences effective radius, and i is an imaginary unit

$r=\sqrt{{(x-x_0)}^2+{(y-y_0)}^2}.$

Near trigpoint, (x, y) the standard point charge model on then is point

${\mathrm{PC}}_{\mathrm{Delta}}={\mathrm{<figure-callout\; id="1"\; label="H"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">H</figure-callout>}}_1+i{\mathrm{<figure-callout\; id="2"\; label="H"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">H</figure-callout>}}_2=\left\{\begin{array}{cc}-\frac{y-y_0}{r}Q-i\frac{x-x_0}{r}Q,& r\&le;\mathrm{<figure-callout\; id="0"\; label="R"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">R</figure-callout>}\\ 0,& r>R\end{array}\right.$

(x wherein ₀, y ₀) be the trigpoint coordinate, other symbolic significances are the same.

It is as follows that model parameter is asked for step:

1, utilizes known weighted least require method, minimize $\underset{(x,y)}{\mathrm{\&Sigma;}}c(x,y){|\mathrm{PR}(x,y)-2\cos \left(\mathrm{\&theta;}(x,y)\right)|}^2$ With $\underset{(x,y)}{\mathrm{\&Sigma;}}c(x,y){|\mathrm{PI}(x,y)-2\sin \left(\mathrm{\&theta;}(x,y)\right)|}^2$ Solve two binary quartic polynomial PR (x, y), PI (x, coefficient y).

Wherein (x, y) be in the fingerprint effective coverage more arbitrarily, (x is that ((x y) is that (x, y) the some field of direction is estimated the angle draw to θ to point for x, certainty level of directional field y) y) to c.

2, solution point charge model parameter

2.1 utilize the multinomial model of obtaining in 1, obtain this model at the value PR at each singular point place (x ₀, y ₀), PI (x ₀, y ₀), utilize cos 2 θ ' (x then ₀, y ₀)=PR (x ₀, y ₀) and sin 2 θ ' (x ₀, y ₀)=PI (x ₀, y ₀) can instead obtain θ ' (x ₀, y ₀), it can be regarded as the anglec of rotation of point charge model with respect to master pattern:

H in the point charge model of then postrotational central point correspondence ₁, H ₂Can be expressed as:

$H_1(x,y)=\frac{y-y_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)-\frac{x-x_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

$H_2(x,y)=\frac{y-y_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)+\frac{x-x_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

H in the charge model of trigpoint correspondence ₁, H ₂For:

$H_1(x,y)=-\frac{y-y_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)-\frac{x-x_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

$H_2(x,y)=-\frac{y-y_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)+\frac{x-x_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

More than respectively to measure implication in four formulas identical with preamble.

2.2 with the point charge model expression formula substitution mixture model formula of trying to achieve, use the electric weight that least square method is obtained each point charge, promptly find the solution following optimization problem:

$\min J=\underset{\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}\left(c(x,y)\right[R(x,y)-\cos \left(2\mathrm{\&theta;}(x,y)\right){]}^2+[I(x,y)-\sin \left(2\mathrm{\&theta;}(x,y)\right){]}^2)$

Ω is the some set that point charge works regional, as shown in Figure 9.

The result that multinomial model is mixed with point charge model and the two is referring to Figure 10.

The figure image intensifying

Algorithm for image enhancement adopts the Gabor filtering method, promptly according to each point field of direction value, carries out filtering with the Gabor wave filter.Effect is as Figure 11-4 after the filtering) shown in.Filtering algorithm is as follows:

1, ask for the spatial domain mask of specifying size:

Gabor wave filter spatial domain expression-form is

$G(x,y,\mathrm{\&theta;})=\exp \left\{\frac{-1}{2}\right[\frac{x^{\&prime;2}}{{\mathrm{\&delta;}}_{x^{\&prime;}}^2}+\frac{y^{\&prime;2}}{{\mathrm{\&delta;}}_{y^{\&prime;}}^2}\left]\right\}\cos \left(2\mathrm{\&pi;}{\mathrm{fx}}^{\&prime;}\right),$ Wherein $\left\{\begin{array}{c}{x}^{\&prime;}=x\sin \mathrm{\&theta;}+y\cos \mathrm{\&theta;}\\ y^{\&prime;}=x\cos \mathrm{\&theta;}-y\sin \mathrm{\&theta;}\end{array};\right.$

Here θ ∈ [0,180) be the field of direction vertical direction of current point, x, y be in the mask each point with respect to the coordinate of mask central point.Each parameter is got δ _x'=δ _y'=5.0, f=0.6, spatial domain mask size is 7 * 7 pixels.Because for identical θ, the spatial domain mask is identical.Finish and store so can be before filtering the spatial domain mask be once asked for, to reduce unnecessary double counting.

2, auto adapted filtering:

For the point in the fingerprint image (i, j), suppose input fingerprint gray level image be I (x, y), θ be (i j) puts the vertical direction of field of direction direction, then uses above-mentioned filter filtering as follows:

$F(i,j)=\frac{{\mathrm{\&Sigma;}}_{x=-w}^w{\mathrm{\&Sigma;}}_{y=-w}^{w}G(x,y,\mathrm{\&theta;})I(i+x,j+y)}{{\mathrm{\&Sigma;}}_{x=-w}^w{\mathrm{\&Sigma;}}_{y=-w}^{w}G(x,y,\mathrm{\&theta;})},$ W=3 wherein;

Ask for a numerical value by following formula then:

$\mathrm{flag}(i,j)=\frac{{\mathrm{\&Sigma;}}_{x=-\frac{L}{2}}^{\frac{L}{2}-D}|F(i+x\cos \mathrm{\&theta;},j+x\sin \mathrm{\&theta;})-F(i+(x+D)\cos \mathrm{\&theta;},j+(x+D)\sin \mathrm{\&theta;})|}{{\mathrm{\&Sigma;}}_{x=-\frac{L}{2}}^{\frac{L}{2}-D}\left(\begin{array}{c}|F(i+x\cos \mathrm{\&theta;},j+x\sin \mathrm{\&theta;})-F(i+(x+D)\cos \mathrm{\&theta;},j+(x+D)\sin \mathrm{\&theta;})|\\ \&times;F[i+(x+\frac{D}{2})\cos \mathrm{\&theta;},j+(x+\frac{D}{2})\sin \mathrm{\&theta;}]\end{array}\right)};$

Wherein L=12 is the statistics zone length, and D=2 is the statistics step-length, carries out crestal line and extracts: if F (i, j)＞(i, j), then (i j) is positioned at paddy (background) to flag, otherwise is positioned at ridge (prospect).

Minutiae point is extracted

Minutiae point can be divided into two kinds, and a kind of is the end points of crestal line, and another kind is the bifurcation point of crestal line, as shown in Figure 2.The method that is based on refinement figure of our employing of the extracting method of minutiae point.Respectively prospect and background are carried out refinement, obtain two refinement figure.Net result is as Figure 11-5,6) shown in.

Concrete thinning method is as follows:

To the fingerprint image after strengthening, we are with its binaryzation (directly select threshold value be 128 get final product).Each some value is 1 or 0,1 expression prospect; 0 expression background.The target of refinement is exactly that to investigate each value be 1 point, value according to its 8 neighborhood determines whether that the point that this is to be investigated is changed to 0 (be about to this point and become background), by traversal several times to full figure, constantly the point with some prospects (value is 1) becomes background dot, thereby reaches the purpose of refinement.

We decide tested point " going " or " staying " according to the different conditions of tested point 8 neighborhoods.The institute of these 8 neighborhood points might the value combination have 2 ⁸=256 kinds (each point can only get 1 or 0).We may be set at a corresponding result of rule with every kind and be " 1 " (reservation) or " 0 " (removal), and the principle that rule is set is the skeleton that keeps former figure.For the fingerprint ridge line refinement, the skeleton of our definition, can be understood as the axis of image, for example rectangular skeleton is the axis on its length direction, the skeleton of circle is its center of circle, the skeleton of ring is the closed curve of similar circle, and the skeleton of straight line is it self, and the skeleton of isolated point also is self.The different application occasion has difference to the definition of skeleton, and we have provided several examples by several examples explanations among Fig. 8, and wherein: (1) can not delete, because it is an internal point, if leave out, has not just had skeleton; (2) can not delete, this is a specific (special) requirements, keeps straight line as far as possible; (3) can not delete, this point is a skeleton, after deleting, changes topological structure; (4) can not delete, because after deleting, originally the part that links to each other disconnects, and changes topological structure; (5) can not delete, because it is the end points of straight line; (6) can delete, this point is not a skeleton; (7) can not delete, this point is a skeleton; (8) can delete, this point is not a skeleton.Our simplified summary once, following criterion is arranged: (1) straight line end points can not be deleted; (2) point that changes topological structure can not be deleted, and for example internal point can not be deleted, isolated point can not be deleted etc.

All situations is summed up according to top example, can obtain 256 rules, it (is exactly an one-dimension array in fact that its result is encoded to a table, mark 0 ~ 255, totally 256 elements), the number that the value of 8 neighborhoods of each tested point is corresponding 0 to 255, with this number as index, the value of correspondence in tabling look-up is if 1 expression keeps; 0 this point of expression is removed (value that is about to this tested point is changed to 0).

Indexing means such as Fig. 7, Aij represent the point in 8 neighborhoods, and index is defined as:

index＝A32×2 ⁰+A31×2 ¹+A21×2 ²+A11×2 ³+A12×2 ⁴+A13×2 ⁵+A23×2 ⁶+A33×2 ⁷

Find corresponding element table[index among the table table according to index value], wherein the span of index value index is at [0,255] interior integer, if table[index] be 1, then keep this point (value is constant); If be 0, then this point is put 0.The table that we adopt as shown in Figure 6.Selecting two from eight examples of Fig. 8 explains as follows:

Fig. 8-2): center (to be measured) point can not be deleted, because:

index＝1×2 ⁰+0×2 ¹+0×2 ²+0×2 ³+1×2 ⁴+0×2 ⁵+1×2 ⁶+0×2 ⁷＝81，

Table[81]=1, can not remove so represent this point.

Fig. 8-8): center (to be measured) point can be deleted, because:

index＝1×2 ⁰+1×2 ¹+0×2 ²+0×2 ³+0×2 ⁴+0×2 ⁵+0×2 ⁶+0×2 ⁷＝3，

Table[3]=0, can remove so represent this point.

We sum up the step of refinement:

The first step provides indexing means, for example according to the method for setting among Fig. 7;

In second step, provide concordance list according to rule, for example according to the table of setting among Fig. 6;

In the 3rd step, traversal full figure all values is 1 point, and computation index judges whether to keep;

In the 4th step, if the 3rd step was not removed any point, then next step otherwise repeated for the 3rd step.

The 5th step, aftertreatment, we will be described in detail below:

The operation of obtaining behind the refinement figure is as follows:

The first step according to refinement figure, is tentatively determined end points (this is as 1 and to have and only have a point in 8 points on every side be 1) in the minutiae point and bifurcation point (basis is as 1 and to have and only have three points in 8 points on every side be 1).

In second step,, minutiae point is carried out aftertreatment along the minutiae point growth:

(a), for end points, if the direction that another end points is arranged in its neighborhood of 12 * 12 is then all removed these two end points with it near (differential seat angle is less than 30 degree);

(b), the adjacent sub crunode that forms annular is coupled together, for a bifurcation, if the direction that another bifurcation is arranged in its neighborhood of 12 * 12 is then removed the both with it near (differential seat angle is less than 30 degree);

(c), remove two end points of some little stub correspondences, for an end points, if along its place crestal line through just meeting another end points within 12 pixels, then two end points are all removed;

In the 3rd step, screen out direction and this field of direction differential seat angle unique point greater than 30 degree

All registered fingerprints are carried out above-mentioned feature extraction operation and deposit the gained feature in database, comprise the coordinate and the direction value of minutiae point, the electric weight of point charge in the coefficient of multinomial model and the point charge model in the mixture model.

Feature relatively

At first the application fingerprint is carried out above-mentioned feature extraction operation, the feature with registered fingerprint in itself and the database compares again.

The minutiae point comparison

The comparison process of minutiae point is divided into minutiae point registration and two steps of minutiae point coupling.

Owing to have rotation and translation between two pieces of fingerprints that are used to compare, must utilize the method compensation rotation and the shifting deviation of minutiae point registration.The method for registering that is based on the Hough conversion that we adopt.Simplicity of explanation is: the minutiae point separately of two fingerprints is constituted two point sets (a M and N minutiae point is respectively arranged) respectively, concentrate from two points and respectively select a minutiae point to be expressed as (x respectively ₁, y ₁, θ ₁) and (x ₂, y ₂, θ ₂), utilize coordinate, direction between them can obtain a translational movement and rotation amount: ${\mathrm{\&Delta;}}_t=\left(\begin{array}{c}{\mathrm{\&Delta;}}_x\\ {\mathrm{\&Delta;}}_y\end{array}\right)=\left(\begin{array}{c}{x}_2-x_1\\ y_2-{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">y</figure-callout>}}_1\end{array}\right);$ Δ _θ＝θ ₂-θ ₁。Travel through all minutiae point to (altogether M * N to), all translations and rotation amount are voted, be i.e. statistics (Δ _x, Δ _y, Δ _θ) number of times that occurs, the highest translation rotation amount of win the vote is exactly the translation rotation amount of final use, writes down number of votes obtained vote simultaneously.

Carry out a little rotation translation transformation according to following formula:

x″＝x′×cos(Δ _θ)-y′×sin(Δ _θ)+Δ _x；

y″＝x′×sin(Δ _θ)-y′×cos(Δ _θ)+Δ _y；

Wherein (x ', y ') be the coordinate before the rotation translation, (x ", y ") is the coordinate after the rotation translation.For the effective coverage, apex coordinate before the difference correspondent transform and the apex coordinate after the conversion; For minutiae point, then distinguish before the corresponding registration and the coordinate behind the registration.

With the public effective coverage that just can calculate after the conversion of two pieces of fingerprint active zones between two pieces of fingerprints (be designated as r, t, and suppose that r is a benchmark, t rotates translation to r).Acquiring method is as follows: good fingerprint effective coverage is respectively R to establish two width of cloth registrations _r, R _t, then public effective coverage is R _c=R _r∩ R _t, R wherein _tBe that effective coverage by the not registration of fingerprint t obtains by the parameter rotation translation of trying to achieve above, the coordinate on 16 summits that has been actually the rotation translation.

For these two pieces good fingerprints of registration, carry out the minutiae point comparison.What finally draw is a number between 0～1, represents the similarity of two pieces of fingerprint minutiaes set.When the distance of two minutiae point in the good fingerprint image of two width of cloth registrations during less than a certain threshold value (being taken as 8 pixels), think that these two somes compares successfully, the match is successful puts counting is added 1.Finally can obtain:

$M_{\mathrm{rt}}=\frac{\mathrm{count}}{\max ({\mathrm{count}}_t,\mathrm{coun}{t}_r)}\&times;\min (\frac{\mathrm{vote}}{\mathrm{Th}},1);$

Wherein count represents to compare successful minutiae point logarithm, count _rThe minutiae point number of expression fingerprint r in the public effective coverage of two width of cloth fingerprints, count _tThe minutiae point number of expression fingerprint t in the public effective coverage of two width of cloth fingerprints.Th is an empirical value, is taken as 12.

The directional diagram comparison

At first directional diagram is aimed at according to alignment methods based on minutiae point.In order to design sorter, need the distance metric between the definition both direction figure based on directional diagram.If O _r, O _tBe respectively the directional diagram of the fingerprint that two width of cloth have aimed at, public effective coverage is R _c(to compare used public domain identical with minutiae point, need not repeat to ask for), then the distance of this two width of cloth directional diagram is:

$S(O_r,O_t)=\frac{1}{\left|R_c\right|}\underset{(i,j)\&Element;R_c}{\mathrm{\&Sigma;}}f(O_r(i,j),O_t(i,j));$

Wherein, | R _c| be public effective coverage area, O _z(i, j) (z=r, t) point that recovers by directional diagram model among the z for directional diagram (i, direction j).Restoration methods is as follows:

At first, according to the formula that provides in the field of direction modeling:

$\left(\begin{array}{c}R(i,j)\\ I(i,j)\end{array}\right)={\mathrm{\&alpha;}}_{\mathrm{PM}}\&CenterDot;\left(\begin{array}{c}\mathrm{PR}(i,j)\\ \mathrm{PI}(i,j)\end{array}\right)+\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{PC}}^{\left(k\right)}\&CenterDot;\left(\begin{array}{c}{H}_1^{\left(k\right)}(i,j)\\ H_2^{\left(k\right)}(i,j)\end{array}\right);$

Calculate point (i, R j) (i, j), I (i, j), wherein each symbolic significance and preamble are identical, recover the field of direction according to following formula then:

F (O _r(i, j), O _t(i, j)) is the function of tolerance differential seat angle.Elect exponential function as: exp (β * (| O _r(x, y)-O _t(x, y) | %90)), wherein % represents to get remainder operation, β=10.

In order to merge with method based on minutiae point, will be apart from being converted to similarity.Similarity=1-S (O _r, O _t).

Decision Fusion

After utilizing minutiae point and directional diagram that fingerprint is compared respectively, can obtain final recognition result by Multiple Classifier Fusion.These two kinds of features have been portrayed fingerprint from different aspects: minutiae point has been portrayed the crestal line distribution characteristics of fingerprint part; Directional diagram has been portrayed the streakline trend of the fingerprint overall situation.Therefore they can constitute the better character representation to fingerprint.At cognitive phase, the sorter of being constructed by their complements each other, and complements one another, for example: based on the identification of directional diagram, can overcome the mistake identification that traditional recognition methods based on minutiae point causes because the very few or false point of minutiae point is too much.We show that fusion can increase substantially the accuracy and the robustness of fingerprint recognition in the experiment on 6616 width of cloth fingerprint databases.The concrete grammar of Decision Fusion is a lot, and the convergence strategy that we adopt is as follows.

If the output of adopting minutiae point and directional diagram to discern is respectively X ₁, X ₂(two numbers), these two identifyings can be seen as two sorters, suppose that they are separate, i.e. X ₁, X ₂Separate, then their joint probability density can be utilized the probability density of two sorters to multiply each other to obtain:

P(X ₁，X ₂|w ₁)＝P(X ₁|w ₁)P(X ₂|w ₁)；

P(X ₁，X ₂|w ₂)＝P(X ₁|w ₂)P(X ₂|w ₂)；

Wherein, P (X ₁, X ₂| w ₁) be illustrated in sample (X ₁, X ₂) (i.e. two comparison results between an input fingerprint and the reference fingerprint) belong to w ₁Probability during class (promptly import fingerprint and reference fingerprint and in fact belong to same finger), P (X ₁, X ₂| w ₂) be illustrated in sample (X ₁, X ₂) belong to w ₂Probability during class (promptly import fingerprint and in fact belong to different fingers) with reference fingerprint.P (X ₁| w ₁) expression belongs to w ₁The fingerprint of class is right, is X with the minutiae point comparison result ₁Probability density, P (X ₂| w ₁) expression belongs to w ₁The fingerprint of class is right, is X with the directional diagram comparison result ₂Probability density, P (X ₁| w ₂) expression belongs to w ₂The fingerprint of class is right, is X with the minutiae point comparison result ₁Probability density, P (X ₂| w ₂) expression belongs to w ₂The fingerprint of class is right, is X with the directional diagram comparison result ₂Probability density.

We can calculate the probability distribution of " real fingerprint comparison " and " false fingerprint comparison ", i.e. P (X by the training sample of some ₁| w ₁), P (X ₂| w ₁), P (X ₁| w ₁) and P (X ₂| w ₂).Method adopts the Parzen window to carry out non-parametric estmation, and window function recommends to adopt Gaussian window, distributes for one dimension, and formula is as follows:

Wherein

N ' is the training sample number, decides on actual amount of data, recommends more than 2000; h _NBe window width, be taken as 0.02; x _iIt is the comparison result of i training sample.Specific algorithm is as follows:

With [0,1] of similarity interval 100 five equilibriums, make the x in the following formula get 0,0.01,0.02 respectively ..., 1 (similarity that actual computation goes out will be similar to according to this precision) is according to the class conditional probability density P (X of following formula computational details point comparison similarity ₁| w ₁), P (X ₁| w ₂) and the class conditional probability density P (X of directional diagram comparison ₂| w ₁) and P (X ₂| w ₂), these four amounts are stored as four arrays that comprise 101 elements in calculating, and with float type (4 byte) storage, need 1616 bytes altogether.Write as file at last, to be used for the on-line decision fusion process of back.Online Decision Fusion only need read corresponding numerical value from file, calculate joint probability density and get final product, and does not need the On-line Estimation probability density.

During online application, obtain the similarity of the similarity of minutiae point comparison between two fingerprints and directional diagram comparison after, utilize the P (X that estimates in advance ₁| w ₁), P (X ₁| w ₂), P (X ₂| w ₁) and P (X ₂| w ₂) obtain P (X ₁, X ₂| w ₁) and P (X ₁, X ₂| w ₂).Sorter below adopting again judges whether these two fingerprints belong to same finger and (promptly belong to w ₁Class or w ₂Class):

$X\&Element;\left\{\begin{array}{cc}{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">w</figure-callout>}}_1& \mathrm{if}(\frac{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2|w_1)}{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2|w_2)}\&GreaterEqual;\mathrm{\&lambda;})\\ {\mathrm{<figure-callout\; id="2"\; label="w"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">w</figure-callout>}}_2& \mathrm{if}(\frac{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2|w_1)}{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2|w_2)}<\mathrm{\&lambda;})\end{array};\right.$

Wherein λ is the likelihood ratio threshold value, requires to select according to reality.Recommend to use 25.

Experiment shows, according to the above-mentioned parameter setting, the false acceptance rate of fingerprint recognition system (soon the fingerprint of other fingers is thought this finger) is generally near 0.01%, and false rejection rate (fingerprint of same finger is thought other fingers) is between 10%-20%, can reduce about 10% when comparing with minutiae point than single.The λ increase can make FAR reduce, and FRR is increased.Should suitably weigh in actual the use.

Description of drawings

Two pieces of different fingerprints of Fig. 1, but minutiae point is compared 25;

Fig. 2 is the composition of many characteristic fingerprints recognition system;

Fig. 3 is the flow process of the many feature extractions of fingerprint;

Fig. 4 is the Sobel operator, 1) and be the horizontal direction operator, 2) be the vertical direction operator.

Fig. 5 is the type of minutiae point and singular point, 1) and be minutiae point, its mid point (1-1) is an end points, point (1-2) is a bifurcation, 2) be the singular point type, point (2-1) is a central point, point (2-2) is a trigpoint;

Fig. 6 is the refinement concordance list;

Fig. 7 sets up the mode synoptic diagram for the refinement index;

Fig. 8 is eight examples of refinement;

Fig. 9 is the point charge model zone of working, and C1, C2 represent two central points, and D1, D2 represent two trigpoints:

Figure 10 is the intermediate result during mixture model calculates, wherein, 1) be the multinomial model fitting result, 2) be the point charge model result, 3) be the result that multinomial model is mixed with point charge model;

Figure 11 is that fingerprint is discerned each step intermediate result, wherein, 1) is original fingerprint figure, 2) be original effective coverage, 3) be the effective coverage of representing with ten hexagons, 4) be to be positioned at the fingerprint image (following each figure is all in the effective coverage) that the effective coverage strengthens, 5) be refinement figure to " ridge " (prospect), 6) be refinement figure to " paddy " (background), 7) be to use the direction field pattern of trying to achieve based on the method for gradient, 8) be the gray level expressing of certainty level of directional field, gray level high more (bright more) represents that this certainty level of directional field is high more, 9) be the direction field pattern that draws with the mixture model match, 10) be the position (for outstanding minutiae point position, fingerprint image done reduction) of minutiae point in fingerprint image of finally trying to achieve.

Figure 12 is simple details of use point and the contrast of the system performance of using many features.

Embodiment

Our invention realizes on common PC computing machine, operating system is not required.

The invention is characterized in that it contains following two stages successively:

Registration phase, the computing machine fingerprint to all registrations under off-line state carries out feature extraction and storage, and this stage is contained following steps successively:

(1). computing machine is carried out initialization:

Set: with following many character representations fingerprint:

Singular point, i.e. field of direction discrete point, it comprises:

Central point: be less than three;

Trigpoint: be less than three;

The X coordinate of singular point, each deposits the Y coordinate with a byte, totally 12 bytes;

Minutiae point comprises its X, Y coordinate and orientation angle, and each minutiae point is with 3 byte representations, and details is counted less than 100, deposits 300 bytes at most in;

Directional diagram uses mixture model, and it contains:

The multinomial model part: with the match of two binary quartic polynomials, n=4, each multinomial coefficient are 25 parameters, amount to 50, with the float type storage of 4 bytes, totally 200 bytes;

The point charge model part, with the storage of float type, unusual when counting less than 6, totally 24 bytes;

The effective coverage chain code is represented border, fingerprint effective coverage with ten hexagons, X, the Y coordinate on 16 summits of storage, totally 32 bytes;

Store above four fingerprint characteristics and take 568 bytes altogether;

In the detection step of fingerprint effective coverage, for the original fingerprint image of having cut apart by 4 * 4 big lattices, when (i j) is gray average I in the grid in the upper left corner with point _Avg(i, j) (i, when j) being in the following ranges, this grid is effectively, is labeled as 1, otherwise invalid, is labeled as 0, that is: with variance Var

Th1＜I _Avg(i, j)＜th2 and simultaneously Var (i, j)＞th3;

Wherein th is a threshold value, th1=20, th2=220, th3=6;

In the step that the field of direction is estimated, the higher limit of field of direction consistency level is T _c=1.5;

In the step of field of direction modeling, the polynomial expression order is 4 in the multinomial model, and central spot point charge model useful effect radius is 10 pixels, and trigpoint place point charge model useful effect radius is 15 pixels;

In the step of figure image intensifying, Gabor wave filter spatial domain expression-form G (x, y, θ) the parameter value δ in _x'=δ _y'=5.0, f=0.6, spatial domain mask size is 7 * 7 pixels;

(2). computing machine is gathered the original image and the storage of all registered fingerprints by getting the finger device;

(3). detect the effective coverage of fingerprint, it contains following steps successively:

(3.1) original image is divided into the grid that size is 4 * 4 pixels;

(3.2) computing machine is calculated as follows that (i j) is gray average I in the grid in the upper left corner with point _Avg(i, j) and variance Var (i, j):

$I_{\mathrm{avg}}(i,j)=\frac{1}{16}\underset{x=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{y=0}{\overset{3}{\mathrm{\&Sigma;}}}I(i+x,j+y);$

$\mathrm{Var}(i,j)=\frac{1}{16}\underset{x=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{y=0}{\overset{3}{\mathrm{\&Sigma;}}}{(I(i+x,j+y)-I_{\mathrm{avg}}(i,j))}^2;$

Wherein (i+x j+y) is (i+x, gradation of image value j+y) to I;

(3.3) computing machine is pressed following formula and is judged whether above-mentioned each grid is effective:

If th1＜I _Avg(i, j)＜th2 and simultaneously Var (i, j)＞th3, then this grid is effective, is labeled as 1, otherwise is 0;

(3.4) eliminated noise is handled:

(3.4.1) above-mentioned image being carried out 3 * 3 filtering, promptly to points all on the image, check 9 points in its 3 * 3 neighborhood, be effectively if having only this point, and other 8 points is invalid, and then this point is judged as noise, and mark changes 0 into; If it is invalid having only this point, and other 8 points are effectively, and then this mark changes 1 into;

(3.4.2) remove " hole " in the effective coverage: promptly line by line to above-mentioned image scanning, fill up all Null Spots between Far Left and the rightmost available point, being labeled as effectively of they; By column scan, fill up topmost and all Null Spots between the available point bottom, being labeled as effectively of they;

Obtained the effective coverage of fingerprint image like this, length and width is respectively 1/4 of former figure;

(3.5) ask for the barycenter (x of real effective coverage ₀, y ₀):

$x_0=\frac{1}{N}\underset{(i,j)\&Element;E}{\mathrm{\&Sigma;}}i,$ ${\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">y</figure-callout>}}_0=\frac{1}{N}\underset{(i,j)\&Element;E}{\mathrm{\&Sigma;}}j;$

Wherein, E represents the effective coverage, and N represents counting in the effective coverage;

(3.6) from barycenter (x ₀, y ₀) locate equal angles ground to around 16 direction injection lines, obtain the intersection point of these rays and effective coverage outer boundary, and store;

(4). use based on the algorithm travel direction field of gradient statistics and estimate that it contains following steps successively:

(4.1) utilize Soble horizontal direction operator S _xWith vertical direction operator S _yAsk for point (x, shade of gray y):

Horizontal direction: $G_x(x,y)=\frac{1}{8}{\mathrm{\&Sigma;}}_{j=y-1}^{y+1}{\mathrm{\&Sigma;}}_{i=x-1}^{x+1}{S}_x(x-i,y-j)I(i,j);$

Vertical direction: $G_y(x,y)=\frac{1}{8}{\mathrm{\&Sigma;}}_{j=y-1}^{y+1}{\mathrm{\&Sigma;}}_{i=x-1}^{x+1}{S}_y(x-i,y-j)I(i,j);$

(i j) is (i, gray-scale value j) to I;

(4.2) fingerprint image being divided into size is

Grid, $\stackrel{-}{W}=7,$ Ask for the local direction θ of each grid correspondence with following formula:

$\mathrm{\&theta;}(i,j)=\frac{1}{2}{\tan }^{-1}\left(\frac{{\mathrm{\&Sigma;}}_{i=1}^{\stackrel{-}{W}}{\mathrm{\&Sigma;}}_{j=1}^{\stackrel{-}{W}}{2G}_x(i,j)G_y(i,j)}{{\mathrm{\&Sigma;}}_{i=1}^{\stackrel{-}{W}}{\mathrm{\&Sigma;}}_{j=1}^{\stackrel{-}{W}}(G_x^2(i,j)-G_y^2(i,j))}\right);$

(4.3) calculated direction field degree of confidence, value are between [0,1], and each point in the field of direction calculates around it The consistance of the field of direction in the neighborhood of size, wherein $\stackrel{-}{W}=7,$ That is:

$c(i,j)=\frac{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}(G_x^2(i+x,j+y)-G_y^2(i+x,j+y))\right)}^2}{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}(G_x^2(i+x,j+y)+G_y^2(i+x,j+y))\right)}^2}$

$+\frac{4\&times;{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}(G_x(i+x,j+y)\&times;G_y(i+x,j+y))\right)}^2}{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{-}{W}/2}^{\stackrel{-}{W}/2}(G_x^2(i+x,j+y)+G_y^2(i+x,j+y))\right)}^2};$

C (i, j) i.e. point (i, certainty level of directional field j) wherein;

(5). on the field of direction, carry out singular point with the Poincare index method and detect:

If: Ψ _x(k) and Ψ _y(k) representing respectively promptly to be asked with set point a little is that radius centered is k horizontal stroke, the ordinate of putting on the circle of 8 pixels, has N on the circle _ψ=25 picture elements;

Then the direction of round adjacent 2 k ', k is changed to:

Wherein: δ (k)=O (Ψ _x(k '), Ψ _y(k '))-O (Ψ _x(k), Ψ _y(k));

k＝(k+1)modN _ψ；

O () represent the field of direction that this fingerprint image k ' or k order [0,180) between; Point (i, j) the Poincare index on can be represented with following formula:

$\mathrm{Poincare}(i,j)=\frac{1}{360}\underset{k=0}{\overset{N_{\mathrm{\&psi;}}-1}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}\left(k\right);$

When the Poincare index was 0, set point was a general point; When the Poincare index was 1/2, set point was a central point; The Poincare index is-1/2 o'clock, and set point is a trigpoint;

The testing result of storage singular point;

(6). set up field of direction mixture model, calculate the model parameter of mixture model, i.e. multinomial coefficient and point charge electric weight;

(6.1) to having a few in the effective coverage, utilize known weighted least require method, minimize respectively $\underset{(x,y)}{\mathrm{\&Sigma;}}c(x,y){|\mathrm{PR}(x,y)-\cos \left(2\mathrm{\&theta;}(x,y)\right)|}^2$ With $\underset{(x,y)}{\mathrm{\&Sigma;}}c(x,y){|\mathrm{PI}(x,y)-\sin \left(2\mathrm{\&theta;}(x,y)\right)|}^2,$ Solve two binary quartic polynomial PR (x, y), PI (x, y) in every coefficient; Wherein (x, y) be in the fingerprint effective coverage more arbitrarily, c (x, y) be point (x, certainty level of directional field y), θ (x, y) be in the step (4) (x, y) the some field of direction is estimated the angle draw;

(6.2) solution point charge model parameter:

(6.2.1) utilize the multinomial model of obtaining in (6.1), obtain this model at each singular point (x ₀, y ₀) the value PR (x that locates ₀, y ₀), PI (x ₀, y ₀), calculate the anglec of rotation θ ' (x of point charge model then according to following formula ₀, y ₀):

Central point (x ₀, y ₀) neighbouring satisfied $r=\sqrt{{(x-x_0)}^2+{(y-y_0)}^2}<R$ Point (x, y) point charge model on is:

$H_1(x,y)=\frac{y-y_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)-\frac{x-x_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

$H_2(x,y)=\frac{y-y_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)+\frac{x-x_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

Q is a central point point charge electric weight to be asked, and R is that given central point point charge influences effective radius; At trigpoint (x ₀, y ₀) neighbouring satisfied $r=\sqrt{{(x-x_0)}^2+{(y-y_0)}^2}<R$ Point (x, y) point charge model on is:

$H_1(x,y)=-\frac{y-y_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)-\frac{x-x_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

$H_2(x,y)=-\frac{y-y_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right)+\frac{x-x_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_0)\right);$

Q is a trigpoint point charge electric weight to be asked, and R is that given trigpoint point charge influences effective radius;

(6.2.2) use the electric weight that least square method is obtained each point charge, promptly find the solution following optimization problem:

$\min J=\underset{\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}\left(c(x,y)\right[R(x,y)-\cos \left(2\mathrm{\&theta;}(x,y)\right){]}^2+[I(x,y)-\sin \left(2\mathrm{\&theta;}(x,y)\right){]}^2);$

Wherein Ω is the work some set in zone of point charge, R (x, y), I (x, y) be mixture model point (x, value y) are asked for according to following formula:

$\left(\begin{array}{c}R(x,y)\\ I(x,y)\end{array}\right)={\mathrm{\&alpha;}}_{\mathrm{PM}}\&CenterDot;\left(\begin{array}{c}\mathrm{PR}(x,y)\\ \mathrm{PI}(x,y)\end{array}\right)+\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{PC}}^{\left(k\right)}\&CenterDot;\left(\begin{array}{c}{H}_1^{\left(k\right)}(x,y)\\ H_2^{\left(k\right)}(x,y)\end{array}\right);$

Wherein PR (x, y), (x is that multinomial model is at point (x, value y), H y) to PI ₁ ^(k)(x, y), H ₂ ^(k)(x is that the point charge model of k singular point correspondence is in point (x, value y) y); α _PC ^(k)(x, y) expression k singular point correspondence point charge model point (computing formula is for x, the weight of y) locating: ${\mathrm{\&alpha;}}_{\mathrm{Pc}}^{\left(k\right)}(x,y)=1-\frac{r^{\left(k\right)}(x,y)}{R^K},$ $r^{\left(k\right)}(x,y)=\min (\sqrt{{(x-x_0^{\left(k\right)})}^2+{(y-y_0^{\left(k\right)})}^2},R^{\left(k\right)}),$ R ^(k)Be the effective radius of the point charge model of k singular point correspondence, (x ₀ ^(k), y ₀ ^(k)) be the coordinate of k singular point; The weight of multinomial model is ${\mathrm{\&alpha;}}_{\mathrm{PM}}(x,y)=\max \{1-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{Pc}}^{\left(k\right)}(x,y),0\};$

(7). strengthen image with the Gabor filtering method, its step is as follows:

(7.1) Gabor wave filter spatial domain expression-form is:

$G(x,y,\mathrm{\&theta;})=\exp \left\{\frac{-1}{2}\right[\frac{x^{\&prime;2}}{{\mathrm{\&delta;}}_{x^{\&prime;}}^2}+\frac{y^{\&prime;2}}{{\mathrm{\&delta;}}_{y^{\&prime;}}^2}\left]\right\}\cos \left(2\mathrm{\&pi;}{\mathrm{fx}}^{\&prime;}\right);$ Wherein $\left\{\begin{array}{c}{x}^{\&prime;}=x\sin \mathrm{\&theta;}+y\cos \mathrm{\&theta;}\\ y^{\&prime;}=x\cos \mathrm{\&theta;}-y\sin \mathrm{\&theta;}\end{array}\right.;$

θ ∈ [0,180) be the field of direction vertical direction of current point, x, y be in the mask each point with respect to the coordinate of mask central point, δ _{X '}=δ _{Y '}=5.0, f=0.6, spatial domain mask size is 7 * 7 pixels;

(7.2) auto adapted filtering:

Suppose input fingerprint gray level image be I (x, y), θ be (i j) puts the vertical direction of field of direction direction, then uses above-mentioned filter filtering as follows:

$F(i,j)=\frac{{\mathrm{\&Sigma;}}_{x=-w}^w{\mathrm{\&Sigma;}}_{y=-w}^{w}G(x,y,\mathrm{\&theta;})I(i+x,j+y)}{{\mathrm{\&Sigma;}}_{x=-w}^w{\mathrm{\&Sigma;}}_{y=-w}^{w}G(x,y,\mathrm{\&theta;})},$ W=3 wherein;

Ask for a numerical value by following formula then:

$\mathrm{flag}(i,j)=\frac{{\mathrm{\&Sigma;}}_{x=-\frac{L}{2}}^{\frac{L}{2}-D}|F(i+x\cos \mathrm{\&theta;},j+x\sin \mathrm{\&theta;})-F(i+(x+D)\cos \mathrm{\&theta;},j+(x+D)\sin \mathrm{\&theta;})|}{{\mathrm{\&Sigma;}}_{x=-\frac{L}{2}}^{\frac{L}{2}-D}\left(\begin{array}{c}|F(i+x\cos \mathrm{\&theta;},j+x\sin \mathrm{\&theta;})-F(i+(x+D)\cos \mathrm{\&theta;},j+(x+D)\sin \mathrm{\&theta;})|\\ \&times;F[i+(x+\frac{D}{2})\cos \mathrm{\&theta;},j+(x+\frac{D}{2})\sin \mathrm{\&theta;}]\end{array}\right)};$ Wherein L=12 is the statistics zone length, and D=2 is the statistics step-length, carries out crestal line and extracts: if F (i, j)＞Flag (i, j), then (i j) is positioned at paddy, i.e. background, otherwise be positioned at ridge, i.e. prospect;

(8). extract and storage based on the minutiae point of refinement figure, it comprises following steps successively:

(8.1) promptly do not change topological structure and do not delete under the prerequisite of straight line end points at the skeleton that keeps former figure, decide tested point " going " or " staying " according to the different conditions that with the tested point is 8 neighborhoods at center, " go " usefulness " 0 " expression, " staying " usefulness " 1 " expression;

(8.2) set up one dimension concordance list table, marked index is 0 ~ 255, and totally 256 elements, each element are got 1 expression and kept, and 0 expression is removed;

(8.2) have a few in the traversal effective coverage, investigate its 8 neighborhood, all permutation and combination are mapped between 0 ~ 255 by following formula:

index＝A32×2 ⁰+A31×2 ¹+A21×2 ²+A11×2 ³

+A12×2 ⁴+A13×2 ⁵+A23×2 ⁶+A33×2 ⁷；

Wherein, Aij represents the value of the point in 8 neighborhoods, is that the element of index is table[index by index value in the search index table then], determine this tested point whether to keep or remove;

(8.4) repeating (8.3) occurs up to the point that is not removed;

(8.5) refinement aftertreatment:

(8.5.1), according to refinement figure, tentatively determine the end points in the minutiae point, promptly this is as 1 and to have and only have a point on every side in 8 points be 1 point; Bifurcation point, promptly this is as 1 and to have and only have three points on every side in 8 points be 1 point;

(8.5.2), along the minutiae point growth, minutiae point is carried out aftertreatment:

(8.5.2.1), for end points, if the direction that another end points is arranged in its neighborhood of 12 * 12 is then all removed these two end points with it near (differential seat angle is less than 30 degree);

(8.5.2.2), the adjacent sub crunode that forms annular is coupled together, for a bifurcation, if the direction that another bifurcation is arranged in its neighborhood of 12 * 12 is then removed the both with it near (differential seat angle is less than 30 degree);

(8.5.2.3), remove two end points of little stub correspondence, for an end points, if along its place crestal line through just meeting another end points within 12 pixels, then two end points are all removed;

(8.5.3), screen out direction and this field of direction differential seat angle unique point greater than 30 degree;

All registered fingerprints are carried out above-mentioned feature extraction operation (1) ~ (8) and deposit the gained feature in database, comprise the coordinate and the direction value of minutiae point, the electric weight of point charge in the coefficient of multinomial model and the point charge model in the mixture model;

Qualify Phase:

(1) ~ (8) identical with registration phase

(9) minutiae point comparison

(9.1) use western node method for registering to compensate rotation and shifting deviation, promptly two fingerprints minutiae point is separately constituted the point set that contains M and N minutiae point separately respectively, concentrate from two points and respectively select a minutiae point to be expressed as (x respectively based on the Hough conversion ₁, y ₁, θ ₁) and (x ₂, y ₂, θ ₂), utilize coordinate, direction between them to obtain a translational movement: ${\mathrm{\&Delta;}}_t=\left(\begin{array}{c}{\mathrm{\&Delta;}}_x\\ {\mathrm{\&Delta;}}_y\end{array}\right)=\left(\begin{array}{c}{x}_2-x_1\\ y_2-{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">y</figure-callout>}}_1\end{array}\right),$ Rotation amount: Δ _θ=θ ₂-θ ₁, it is right to minutiae point to travel through all M * N, statistics (Δ _x, Δ _y, Δ _θ) number of times that occurs, the highest translation rotation amount of win the vote is exactly the translation rotation amount of final use, writes down number of votes obtained vote simultaneously;

The coordinate transform of using below can be realized by following formula:

x″＝x′×cos(Δ _θ)-y′×sin(Δ _θ)+Δ _x；

y″＝x′×sin(Δ _θ)-y′×cos(Δ _θ)+Δ _y；

Wherein (x ', y ') be the coordinate before the rotation translation, (x ", y ") is the coordinate after the rotation translation;

(9.2) extract public effective coverage:

Effective coverage behind note registered fingerprint r and the application fingerprint t registration is respectively R _r, R _t, establishing with r is benchmark, and t rotates translation to r, and then public effective coverage is R _c=R _r∩ R _t, R wherein _tBeing to be obtained by the parameter rotation translation that try to achieve by (9.1) effective coverage of fingerprint t, specifically is to realize by the coordinate on formula rotation translation ten hexagonal 16 summits in (9.1);

(9.3) by the coordinate of all minutiae point among formula rotation translation fingerprint t in (9.1), and with its with fingerprint r in all minutiae point compare, it successfully is logarithm apart from less than the minutiae point of 8 pixels that record compare;

(9.4) calculated fingerprint r, the similarity M of t minutiae point set _Rt, 0＜M _Rt＜1:

$M_{\mathrm{rt}}=\frac{\mathrm{count}}{\max ({\mathrm{count}}_t,{\mathrm{count}}_r)}\&times;\min (\frac{\mathrm{vote}}{\mathrm{Th}},1);$

Wherein count represents to compare successful minutiae point logarithm, count _rThe minutiae point number of expression fingerprint r in the public effective coverage of two width of cloth fingerprints, count _tThe minutiae point number of expression fingerprint t in the public effective coverage of two width of cloth fingerprints, Th is an empirical value, is taken as 12;

(10) directional diagram comparison:

(10.1) according to the formula that provides in (6):

$\left(\begin{array}{c}R(x,y)\\ I(x,y)\end{array}\right)={\mathrm{\&alpha;}}_{\mathrm{PM}}\&CenterDot;\left(\begin{array}{c}\mathrm{PR}(x,y)\\ \mathrm{PI}(x,y)\end{array}\right)+\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{PC}}^{\left(k\right)}\&CenterDot;\left(\begin{array}{c}{H}_1^{\left(k\right)}(x,y)\\ H_2^{\left(k\right)}(x,y)\end{array}\right);$

Calculate point (x, R y) (x, y), I (x, y), wherein each symbolic significance and (6) are identical, recover directional diagram according to following formula then:

The aligning of the parameter travel direction figure that draws according to the minutiae point registration, the coordinate that is about to each point in the directional diagram of fingerprint t carries out conversion according to the parameter of determining in (9.1) according to the formula that (9.1) provide;

(10.2) establish O _r, O _tBe respectively the directional diagram of the fingerprint that two width of cloth have aimed at, the effective coverage is R in (9.2) _c, need not repeat to ask for;

(10.3) be calculated as follows two width of cloth directional diagram O _r, O _tDistance:

$S(O_r,O_t)=\frac{1}{\left|R_c\right|}\underset{(x,y)\&Element;R_c}{\mathrm{\&Sigma;}}f(O_r(x,y),O_t(x,y));$

Wherein, | R _c| be public effective coverage area, O _z(x, y) (z=r t) is directional diagram z mid point (x, direction y), f (O _r(x, y), O _t(x, y)) is the function of tolerance differential seat angle, elects exponential function as: exp (β * (| O _r(x, y)-O _t(x, y) | %90)), wherein % represents to get remainder operation, β=10;

(10.4) similarity=1-S (O _r, O _t);

(11). carry out Decision Fusion with sorter:

At first define w ₁In fact the class incident belongs to the situation of same finger for input fingerprint and reference fingerprint; w ₂In fact the class incident belongs to the incident of different fingers for input fingerprint and reference fingerprint;

(11.1) probability density estimation procedure:

Order:

P (X ₁| w ₁) expression: belong to w ₁The fingerprint of class is right, is X with the minutiae point comparison result ₁Probability density;

P (X ₂| w ₁) expression: belong to w ₁The fingerprint of class is right, is X with the directional diagram comparison result ₂Probability density;

P (X ₁| w ₂) expression: belong to w ₂The fingerprint of class is right, is X with the minutiae point comparison result ₁Probability density;

P (X ₂| w ₂) expression: belong to w ₂The fingerprint of class is right, is X with the directional diagram comparison result ₂Probability density;

P (X ₁| w ₁), P (X ₂| w ₁), P (X ₁| w ₂) and P (X ₂| w ₂) use Represent, and ask for following formula:

Wherein

N ' is the training sample number, the number of promptly corresponding comparison result, and four groups of data bulks are recommended all more than 2000, and these comparisons are that off-line carries out; h _NBe window width, be taken as 0.02; X gets 0,0.01,0.02 respectively ..., 1, soon interval [0,1] 100 five equilibriums of similarity calculate P (X according to following formula respectively ₁| w ₁), P (X ₂| w ₁), P (X ₁| w ₂) and P (X ₂| w ₂), these four amounts are stored as four arrays that comprise 101 elements in calculating, and with float type (4 byte) storage, need 1616 bytes altogether; Write as file at last;

(11.2) on-line decision process:

Result (the X that the result who draws according to minutiae point and directional diagram comparison compares ₁', X ₂'), from the file of storage probability density, read corresponding numerical value, calculate joint probability density:

P(X ₁′，X ₂′|w ₁)＝P(X ₁′|w ₁)P(X ₂′|w ₁)，

P(X ₁′，X ₂′|w ₂)＝P(X ₁′|w ₂)P(X ₂′|w ₂)，

Use following decision rule to make a strategic decision:

If $\frac{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1^{\&prime;},{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2^{\&prime;}|w_1)}{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1^{\&prime;},{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2^{\&prime;}|w_2)}\&GreaterEqual;\mathrm{\&lambda;},$ Judge that then fingerprint belongs to same finger;

If $\frac{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1^{\&prime;},{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2^{\&prime;}|w_1)}{P({\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_1^{\&prime;},{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20041000932800321.png,C20041000932800361.png"\; state="\{\{state\}\}">X</figure-callout>}}_2^{\&prime;}|w_2)}<\mathrm{\&lambda;}$ Judge that then fingerprint belongs to different fingers;

Wherein λ is a decision-making value, selects 25 for use.

Claims (1)

1. the many characteristic recognition methods of fingerprint is characterized in that, it contains following two stages successively:

Registration phase, the computing machine fingerprint to all registrations under off-line state carries out feature extraction and storage, and this stage is contained following steps successively:

(1). computing machine is carried out initialization:

Set: with following many character representations fingerprint:

Singular point, i.e. field of direction discrete point, it comprises:

Central point: be less than three;

Trigpoint: be less than three;

The X coordinate of singular point, each deposits the Y coordinate with a byte, totally 12 bytes;

Minutiae point comprises its X, Y coordinate and orientation angle, and each minutiae point is with 3 byte representations, and details is counted less than 100, deposits 300 bytes at most in;

Directional diagram uses mixture model, and it contains:

The multinomial model part: with the match of two binary quartic polynomials, n=4, each multinomial coefficient are 25 parameters, amount to 50, with the float type storage of 4 bytes, totally 200 bytes;

The point charge model part, with the storage of float type, unusual when counting less than 6, totally 24 bytes;

The effective coverage chain code is represented border, fingerprint effective coverage with ten hexagons, X, the Y coordinate on 16 summits of storage, totally 32 bytes;

Store above four fingerprint characteristics and take 568 bytes altogether;

In the detection step of fingerprint effective coverage, for the original fingerprint image of having cut apart by 4 * 4 big lattices, when (i j) is gray average I in the grid in the upper left corner with point _Avg(i, j) (i, when j) being in the following ranges, this grid is effectively, is labeled as 1, otherwise invalid, is labeled as 0, that is: with variance Var

Th1＜I _Avg(i, j)＜th2 and simultaneously Var (i, j)＞th3;

Wherein th is a threshold value, th1=20, th2=220, th3=6;

In the step that the field of direction is estimated, the higher limit of field of direction consistency level is T _c=1.5;

In the step of field of direction modeling, the polynomial expression order is 4 in the multinomial model, and central spot point charge model useful effect radius is 10 pixels, and trigpoint place point charge model useful effect radius is 15 pixels;

In the step of figure image intensifying, Gabor wave filter spatial domain expression-form G (x, y, θ) the parameter value δ in _{X '}=δ _{Y '}=5.0, f=0.6, spatial domain mask size is 7 * 7 pixels;

(2). computing machine is gathered the original image and the storage of all registered fingerprints by getting the finger device;

(3). detect the effective coverage of fingerprint, it contains following steps successively:

(3.1) original image is divided into the grid that size is 4 * 4 pixels;

(3.2) computing machine is calculated as follows that (i j) is gray average I in the grid in the upper left corner with point _Avg(i, j) and variance Var (i, j):

$I_{\mathrm{avg}}(i,j)=\frac{1}{16}\underset{x=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{y=0}{\overset{3}{\mathrm{\&Sigma;}}}I(i+x,j+y);$

$\mathrm{Var}(i,j)=\frac{1}{16}\underset{x=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{y=0}{\overset{3}{\mathrm{\&Sigma;}}}{(I(i+x,j+y)-I_{\mathrm{avg}}(i,j))}^2;$

Wherein (i+x j+y) is (i+x, gradation of image value j+y) to I;

(3.3) computing machine is pressed following formula and is judged whether above-mentioned each grid is effective:

If th1＜I _Avg(i, j)＜th2 and simultaneously Var (i, j)＞th3, then this grid is effective, is labeled as 1, otherwise is 0;

(3.4) eliminated noise is handled:

(3.4.1) above-mentioned image being carried out 3 * 3 filtering, promptly to points all on the image, check 9 points in its 3 * 3 neighborhood, be effectively if having only this point, and other 8 points is invalid, and then this point is judged as noise, and mark changes 0 into; If it is invalid having only this point, and other 8 points are effectively, and then this mark changes 1 into;

(3.4.2) remove " hole " in the effective coverage: promptly line by line to above-mentioned image scanning, fill up all Null Spots between Far Left and the rightmost available point, being labeled as effectively of they; By column scan, fill up topmost and all Null Spots between the available point bottom, being labeled as effectively of they;

Obtained the effective coverage of fingerprint image like this, length and width is respectively 1/4 of former figure;

(3.5) ask for the barycenter (x of real effective coverage ₀, y ₀):

$x_0=\frac{1}{N}\underset{(i,j)\&Element;E}{\mathrm{\&Sigma;}}i,$ $y_0=\frac{1}{N}\underset{(i,j)\&Element;E}{\mathrm{\&Sigma;}}j;$

Wherein, E represents the effective coverage, and N represents counting in the effective coverage;

(3.6) from barycenter (x ₀, y ₀) locate equal angles ground to around 16 direction injection lines, obtain the intersection point of these rays and effective coverage outer boundary, and store;

(4). use based on the algorithm travel direction field of gradient statistics and estimate that it contains following steps successively:

(4.1) utilize Soble horizontal direction operator S _xWith vertical direction operator S _yAsk for point (x, shade of gray y):

Horizontal direction: $G_x(x,y)=\frac{1}{8}{\mathrm{\&Sigma;}}_{j=y-1}^{y+1}{\mathrm{\&Sigma;}}_{i=x-1}^{x+1}{S}_x(x-i,y-j)I(i,j);$

Vertical direction: $G_y(x,y)=\frac{1}{8}{\mathrm{\&Sigma;}}_{j=y-1}^{y+1}{\mathrm{\&Sigma;}}_{i=x-1}^{x+1}{S}_y(x-i,y-j)I(i,j);$

(i j) is (i, gray-scale value j) to I;

(4.2) fingerprint image being divided into size is Grid, $\stackrel{\&OverBar;}{W}=7,$ Ask for the local direction θ of each grid correspondence with following formula:

$\mathrm{\&theta;}(i,j)=\frac{1}{2}{\tan }^{-1}\left(\frac{{\mathrm{\&Sigma;}}_{i=1}^{\stackrel{\&OverBar;}{W}}{\mathrm{\&Sigma;}}_{j=1}^{\stackrel{\&OverBar;}{W}}2G_x(i,j)G_y(i,j)}{{\mathrm{\&Sigma;}}_{i=1}^{\stackrel{\&OverBar;}{W}}{\mathrm{\&Sigma;}}_{j=1}^{\stackrel{\&OverBar;}{W}}(G_x^2(i,j)-G_y^2(i,j))}\right);$

(4.3) calculated direction field degree of confidence, value are between [0,1], and each point in the field of direction calculates around it The consistance of the field of direction in the neighborhood of size, wherein $\stackrel{\&OverBar;}{W}=7,$ That is:

$c(i,j)=\frac{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x^2(i+x,j+y)-G_y^2(i+x,j+y))\right)}^2}{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x^2(i+x,j+y)+G_y^2(i+x,j+y))\right)}^2}$

$+\frac{{4\&times;\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x(i+x,j+y)\&times;G_y(i+x,j+y))\right)}^2}{{\left({\mathrm{\&Sigma;}}_{x=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}{\mathrm{\&Sigma;}}_{y=-\stackrel{\&OverBar;}{W}/2}^{\stackrel{\&OverBar;}{W}/2}(G_x^2(i+x,j+y)+G_y^2(i+x,j+y))\right)}^2};$

C (i, j) i.e. point (i, certainty level of directional field j) wherein;

(5). on the field of direction, carry out singular point with the Poincare index method and detect:

If: Ψ _x(k) and Ψ _y(k) representing respectively promptly to be asked with set point a little is that radius centered is k horizontal stroke, the ordinate of putting on the circle of 8 pixels, has N on the circle _Ψ=25 picture elements;

Then the direction of round adjacent 2 k ', k is changed to:

Wherein: δ (k)=O (Ψ _x(k '), Ψ _y(k '))-O (Ψ _x(k), Ψ _y(k));

k′＝(k+1)modN _Ψ；

O () represent the field of direction that this fingerprint image k ' or k order [0,180) between; Point (i, j) the Poincare index on can be represented with following formula:

$\mathrm{Poincare}(i,j)=\frac{1}{360}\underset{k=0}{\overset{N_{\mathrm{\&psi;}}-1}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}\left(k\right);$

When the Poincare index was 0, set point was a general point; When the Poincare index was 1/2, set point was a central point; The Poincare index is-1/2 o'clock, and set point is a trigpoint;

The testing result of storage singular point;

(6). set up field of direction mixture model, calculate the model parameter of mixture model, i.e. multinomial coefficient and point charge electric weight;

(6.1) to having a few in the effective coverage, utilize known weighted least require method, minimize respectively $\underset{(x,y)}{\mathrm{\&Sigma;}}c(x,y)|\mathrm{PR}(x,y)-\cos \left(2\mathrm{\&theta;}(x,y)\right){|}^2$ With $\underset{(x,y)}{\mathrm{\&Sigma;}}c(x,y)|\mathrm{PI}(x,y)-\sin \left(2\mathrm{\&theta;}(x,y)\right){|}^2,$ Solve two binary quartic polynomial PR (x, y), PI (x, y) in every coefficient; Wherein (x, y) be in the fingerprint effective coverage more arbitrarily, c (x, y) be point (x, certainty level of directional field y), θ (x, y) be in the step (4.2) (x, y) the some field of direction is estimated the angle draw;

(6.2) solution point charge model parameter:

(6.2.1) utilize the multinomial model of obtaining in (6.1), obtain this model at each singular point (x ₀, y ₀) the value PR (x that locates ₀, y ₀), PI (x ₀, y ₀), calculate the anglec of rotation θ ' (x of point charge model then according to following formula ₀, y ₀):

Central point (x ₀, y ₀) neighbouring satisfied $r=\sqrt{{(x-x_0)}^2+{(y-y_0)}^2}<R$ Point (x, y) point charge model on is:

$H_1(x,y)=\frac{y-y_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right)-\frac{x-x_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right);$

$H_2(x,y)=\frac{y-y_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right)+\frac{x-x_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right);$

Q is a central point point charge electric weight to be asked, and R is given central point point charge effective radius;

Trigpoint (x ₀, y ₀) neighbouring satisfied $r=\sqrt{{(x-x_0)}^2+{(y-y_0)}^2}<R$ Point (x, y) charge model on is:

$H_1(x,y)=-\frac{y-y_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right)-\frac{x-x_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right);$

$H_2(x,y)=-\frac{y-y_0}{r}\&times;Q\&times;\sin \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right)+\frac{x-x_0}{r}\&times;Q\&times;\cos \left({\mathrm{\&theta;}}^{\&prime;}(x_0,y_0)\right);$

Q is a trigpoint point charge electric weight to be asked, and R is that given trigpoint point charge influences effective radius;

(6.2.2) use the electric weight that least square method is obtained each point charge, promptly find the solution following optimization problem:

$\min J=\underset{\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}\left(c(x,y)\right[R(x,y)-\cos \left(2\mathrm{\&theta;}(x,y)\right){]}^2+[I(x,y)-\sin \left(2\mathrm{\&theta;}(x,y)\right){]}^2);$

Wherein Ω is the work some set in zone of point charge, R (x, y), I (x, y) be mixture model point (x, value y) are asked for according to following formula:

$\left(\begin{array}{c}R(x,y)\\ I(x,y)\end{array}\right)={\mathrm{\&alpha;}}_{\mathrm{PM}}\&CenterDot;\left(\begin{array}{c}\mathrm{PR}(x,y)\\ \mathrm{PI}(x,y)\end{array}\right)+\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{PC}}^{\left(k\right)}\&CenterDot;\left(\begin{array}{c}{H}_1^{\left(k\right)}(x,y)\\ H_2^{\left(k\right)}(x,y)\end{array}\right);$

Wherein PR (x, y), (x is that multinomial model is at point (x, value y), H y) to PI ₁ ^(k)(x, y), H ₂ ^(k)(x is that the point charge model of k singular point correspondence is in point (x, value y) y); α _PC ^(k)(x, y) expression k singular point correspondence point charge model point (computing formula is for x, the weight of y) locating:

${\mathrm{\&alpha;}}_{\mathrm{Pc}}^{\left(k\right)}(x,y)=1-\frac{r^{\left(k\right)}(x,y)}{R^K},$ $r^{\left(k\right)}(x,y)=\min (\sqrt{{(x-x_0^{\left(k\right)})}^2+{(y-y_0^{\left(k\right)})}^2},R^{\left(k\right)}),$ R ^(k)Be the effective radius of the point charge model of k singular point correspondence, (x ₀ ^(k), y ₀ ^(k)) be the coordinate of k singular point; The weight of multinomial model is ${\mathrm{\&alpha;}}_{\mathrm{PM}}(x,y)=\max \{1-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{Pc}}^{\left(k\right)}(x,y),0\};$

(7). strengthen image with the Gabor filtering method, its step is as follows:

(7.1) Gabor wave filter spatial domain expression-form is:

$G(x,y,\mathrm{\&theta;})=\exp \left\{\frac{-1}{2}\right[\frac{x^{\&prime;2}}{{\mathrm{\&delta;}}_{x^{\&prime;}}^2}+\frac{y^{\&prime;2}}{{\mathrm{\&delta;}}_{y^{\&prime;}}^2}\left]\right\}\cos \left(2\mathrm{\&pi;}{\mathrm{fx}}^{\&prime;}\right);$ Wherein $\left\{\begin{array}{c}{x}^{\&prime;}=x\sin \mathrm{\&theta;}+y\cos \mathrm{\&theta;}\\ y^{\&prime;}=x\cos \mathrm{\&theta;}-y\sin \mathrm{\&theta;}\end{array}\right.;$

θ ∈ [0,180) be the field of direction vertical direction of current point, x, y be in the mask each point with respect to the coordinate of mask central point, δ _{X '}=δ _{Y '}=5.0, f=0.6, spatial domain mask size is 7 * 7 pixels;

(7.2) auto adapted filtering:

Suppose input fingerprint gray level image be I (x, y), θ be (i j) puts the vertical direction of field of direction direction, then uses above-mentioned filter filtering as follows:

$F(i,j)=\frac{{\mathrm{\&Sigma;}}_{x=-w}^w{\mathrm{\&Sigma;}}_{y=-w}^{w}G(x,y,\mathrm{\&theta;})I(i+x,j+y)}{{\mathrm{\&Sigma;}}_{x=-w}^w{\mathrm{\&Sigma;}}_{y=-w}^{w}G(x,y,\mathrm{\&theta;})},$ W=3 wherein;

Ask for a numerical value by following formula then:

$\mathrm{flag}(i,j)=\frac{{\mathrm{\&Sigma;}}_{x=-\frac{L}{2}}^{\frac{L}{2}-D}|F(i+x\cos \mathrm{\&theta;},j+x\sin \mathrm{\&theta;})-F(i+(x+D)\cos \mathrm{\&theta;},j+(x+D)\sin \mathrm{\&theta;})|}{{\mathrm{\&Sigma;}}_{x=-\frac{L}{2}}^{\frac{L}{2}-D}\left(\begin{array}{c}|F(i+x\cos \mathrm{\&theta;},j+x\sin \mathrm{\&theta;})-F(i+(x+D)\cos \mathrm{\&theta;},j+(x+D)\sin \mathrm{\&theta;})|\\ \&times;F[i+(x+\frac{D}{2})\cos \mathrm{\&theta;},j+(x+\frac{D}{2})\sin \mathrm{\&theta;}]\end{array}\right)};$

Wherein L=12 is the statistics zone length, and D=2 is the statistics step-length, carries out crestal line and extracts: if

F (i, j)＞flag (i, j), then (i j) is positioned at paddy, i.e. background, otherwise be positioned at ridge, i.e. prospect;

(8). extract and storage based on the minutiae point of refinement figure, it comprises following steps successively:

(8.1) promptly do not change topological structure and do not delete under the prerequisite of straight line end points at the skeleton that keeps former figure, decide tested point " going " or " staying " according to the different conditions that with the tested point is 8 neighborhoods at center, " go " usefulness " 0 " expression, " staying " usefulness " 1 " expression;

(8.2) set up one dimension concordance list table, marked index is 0 ~ 255, and totally 256 elements, each element are got 1 expression and kept, and 0 expression is removed;

(8.2) have a few in the traversal effective coverage, investigate its 8 neighborhood, all permutation and combination are mapped between 0 ~ 255 by following formula:

index＝A32×2 ⁰+A31×2 ¹+A21×2 ²+A11×2 ³+A12×2 ⁴+A13×2 ⁵+A23×2 ⁶+A33×2 ⁷；

Wherein, Aij represents the value of the point in 8 neighborhoods, is that the element of index is table[index by index value in the search index table then], determine this tested point whether to keep or remove;

(8.4) repeating (8.3) occurs up to the point that is not removed;

(8.5) refinement aftertreatment:

(8.5.1), according to refinement figure, tentatively determine the end points in the minutiae point, promptly this is as 1 and to have and only have a point on every side in 8 points be 1 point; Bifurcation point, promptly this is as 1 and to have and only have three points on every side in 8 points be 1 point;

(8.5.2), along the minutiae point growth, minutiae point is carried out aftertreatment:

(8.5.2.1), for end points, if the direction that another end points is arranged in its neighborhood of 12 * 12 is then all removed these two end points with it near (differential seat angle is less than 30 degree);

(8.5.2.2), the adjacent sub crunode that forms annular is coupled together, for a bifurcation, if the direction that another bifurcation is arranged in its neighborhood of 12 * 12 is then removed the both with it near (differential seat angle is less than 30 degree);

(8.5.2.3), remove two end points of little stub correspondence, for an end points, if along its place crestal line through just meeting another end points within 12 pixels, then two end points are all removed;

(8.5.3), screen out direction and this field of direction differential seat angle unique point greater than 30 degree;

All registered fingerprints are carried out above-mentioned feature extraction operation (1) ~ (8) and deposit the gained feature in database, comprise the coordinate and the direction value of minutiae point, the electric weight of point charge in the coefficient of multinomial model and the point charge model in the mixture model;

Qualify Phase:

(1) ~ (8) identical with registration phase

(9) minutiae point comparison

(9.1) use minutiae point method for registering to compensate rotation and shifting deviation, promptly two fingerprints minutiae point is separately constituted the point set that contains M and N minutiae point separately respectively, concentrate from two points and respectively select a minutiae point to be expressed as (x respectively based on the Hough conversion ₁, y ₁, θ ₁) and (x ₂, y ₂, θ ₂), utilize coordinate, direction between them to obtain a translational movement: ${\mathrm{\&Delta;}}_t=\left(\begin{array}{c}{\mathrm{\&Delta;}}_x\\ {\mathrm{\&Delta;}}_y\end{array}\right)=\left(\begin{array}{c}{x}_2-x_1\\ y_2-y_1\end{array}\right),$ Rotation amount: Δ _θ=θ ₂-θ ₁, it is right to minutiae point to travel through all M * N, statistics (Δ _x, Δ _y, Δ _θ) number of times that occurs, the highest translation rotation amount of win the vote is exactly the translation rotation amount of final use, writes down number of votes obtained vote simultaneously;

The coordinate transform of using below can be realized by following formula:

x″＝x′×cos(Δ _θ)-y′×sin(Δ _θ)+Δ _x；

y″＝x′×sin(Δ _θ)-y′×cos(Δ _θ)+Δ _y；

Wherein (x ', y ') be the coordinate before the rotation translation, (x ", y ") is the coordinate after the rotation translation;

(9.2) extract public effective coverage:

Effective coverage behind note registered fingerprint r and the application fingerprint t registration is respectively R _r, R _l, establishing with r is benchmark, and t rotates translation to r, and then public effective coverage is $R_c=R_r\&cap;R_t,$ R wherein _tBe by the effective coverage of fingerprint t by

The parameter rotation translation of (9.1) trying to achieve obtains, and specifically is to realize by the coordinate on formula rotation translation ten hexagonal 16 summits in (9.1);

(9.3) by the coordinate of all minutiae point among formula rotation translation fingerprint t in (9.1), and with its with fingerprint r in all minutiae point compare, it successfully is logarithm apart from less than the minutiae point of 8 pixels that record compare;

(9.4) calculated fingerprint r, the similarity M of t minutiae point set _Rt, 0＜M _Rt＜1:

$M_{\mathrm{rt}}=\frac{\mathrm{count}}{\max (\mathrm{coun}{t}_t,\mathrm{coun}{t}_r)}\&times;\min (\frac{\mathrm{vote}}{\mathrm{Th}},1);$

Wherein count represents to compare successful minutiae point logarithm, count _rThe minutiae point number of expression fingerprint r in the public effective coverage of two width of cloth fingerprints, count _tThe minutiae point number of expression fingerprint t in the public effective coverage of two width of cloth fingerprints, TH is an empirical value, is taken as 12;

(10) directional diagram comparison:

(10.1) according to the formula that provides in (6):

$\left(\begin{array}{c}R(x,y)\\ I(x,y)\end{array}\right)={\mathrm{\&alpha;}}_{\mathrm{PM}}\&CenterDot;\left(\begin{array}{c}\mathrm{PR}(x,y)\\ \mathrm{PI}(x,y)\end{array}\right)+\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{PC}}^{\left(k\right)}\&CenterDot;\left(\begin{array}{c}{H}_1^{\left(k\right)}(x,y)\\ H_2^{\left(k\right)}(x,y)\end{array}\right);$

Calculate point (x, R y) (x, y), I (x, y), wherein each symbolic significance and (6) are identical, recover directional diagram according to following formula then:

The aligning of the parameter travel direction figure that draws according to the minutiae point registration, the coordinate that is about to each point in the directional diagram of fingerprint t carries out conversion according to the parameter of determining in (9.1) according to the formula that (9.1) provide;

(10.2) establish O _r, O _tBe respectively the directional diagram of the fingerprint that two width of cloth have aimed at, the effective coverage is R in (9.2) _c, need not repeat to ask for;

(10.3) be calculated as follows two width of cloth directional diagram O _r, O _tDistance:

$S(O_r,O_t)=\frac{1}{\left|R_c\right|}\underset{(x,y)\&Element;R_c}{\mathrm{\&Sigma;}}f(O_r(x,y),O_t(x,y));$

Wherein, | R _c| be public effective coverage area, O _z(x, y) (z=r t) is directional diagram z mid point (x, direction y), f (O _r(x, y), O _t(x, y)) is the function of tolerance differential seat angle, elects exponential function as: exp (β * (| O _r(x, y)-O _t(x, y) | %90)), wherein % represents to get remainder operation, β=10; (10.4) similarity=1-S (O _r, O _t);

(11). carry out Decision Fusion with sorter:

At first define w ₁In fact the class incident belongs to the situation of same finger for input fingerprint and reference fingerprint; w ₂In fact the class incident belongs to the incident of different fingers for input fingerprint and reference fingerprint;

(11.1) probability density estimation procedure:

Order;

P (X ₁| w ₁) expression: belong to w ₁The fingerprint of class is right, is X with the minutiae point comparison result ₁Probability density;

P (X ₂| w ₁) expression: belong to w ₁The fingerprint of class is right, is X with the directional diagram comparison result ₂Probability density;

P (X ₁| w ₂) expression: belong to w ₂The fingerprint of class is right, is X with the minutiae point comparison result ₁Probability density;

P (X ₂| w ₂) expression: belong to w ₂The fingerprint of class is right, is X with the directional diagram comparison result ₂Probability density;

P (X ₁| w ₁), P (X ₂| w ₁), P (X ₁| w ₂) and P (X ₂| w ₂) use

Expression, ask for following formula:

Wherein

N ' is the training sample number, the number of promptly corresponding comparison result, and four groups of data bulks are recommended all more than 2000, and these comparisons are that off-line carries out; h _NBe window width, be taken as 0.02; X gets 0,0.01,0.02 respectively ..., 1, soon interval [0,1] 100 five equilibriums of similarity calculate P (X according to following formula respectively ₁| w ₁), P (X ₂| w ₁), P (X ₁| w ₂) and P (X ₂| w ₂), these four amounts are stored as four arrays that comprise 101 elements in calculating, and with float type (4 byte) storage, need 1616 bytes altogether; Write as file at last;

(11.2) on-line decision process:

Result (the X that the result who draws according to minutiae point and directional diagram comparison compares ₁', X ₂'), from the file of storage probability density, read corresponding numerical value, calculate joint probability density:

P(X ₁′，X ₂′|w ₁)＝P(X ₁′|w ₁)P(X ₂′|w ₁)，

P(X ₁′，X ₂′|w ₂)＝P(X ₁′|w ₂)P(X ₂′|w ₂)，

Use following decision rule to make a strategic decision:

If $\frac{P(X_1^{\&prime;},X_2^{\&prime;}|w_1)}{P(X_1^{\&prime;},X_2^{\&prime;}|w_2)}\&GreaterEqual;\mathrm{\&lambda;},$ Judge that then fingerprint belongs to same finger;

If $\frac{P(X_1^{\&prime;},X_2^{\&prime;}|w_1)}{P(X_1^{\&prime;},X_2^{\&prime;}|w_2)}<\mathrm{\&lambda;},$ Judge that then fingerprint belongs to different fingers;

Wherein λ is a decision-making value, selects 25 for use.

Images