CN104866815B - Fingerprint epipole accurate positioning method based on image spatial feature - Google Patents
Fingerprint epipole accurate positioning method based on image spatial feature Download PDFInfo
- Publication number
- CN104866815B CN104866815B CN201510196807.1A CN201510196807A CN104866815B CN 104866815 B CN104866815 B CN 104866815B CN 201510196807 A CN201510196807 A CN 201510196807A CN 104866815 B CN104866815 B CN 104866815B
- Authority
- CN
- China
- Prior art keywords
- matrix
- epipole
- submodule
- streakline
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The fingerprint epipole accurate positioning method based on image spatial feature that the invention discloses a kind of.Specific step is as follows by the present invention:The Pixel-level field of direction of fingerprint image is found out with gradient method;With Poincare indexing method in the field of directionIn extract each thick epipole position;The binary image of fingerprint image is calculated with Short Time Fourier Transform method;By the field of direction, thick epipole position and binary imageIt inputs bifurcated and polymerize extraction module, extract near thick epipole positionAll bifurcations in rectangular extentAnd congruent point, and it is stored in matrixIn;By matrixDenoising module is input to remove matrixAll pseudo- bifurcation, pseudo- congruent point, obtain matrix;The matrix that will be obtained after denoisingIt is input to arc extraction module and obtains all arcs, and obtained arc is stored in matrixIn;By matrixIt is input to pinpoint module and obtains the final result of epipole positioning.Present invention utilizes the spatial information (si) of print image, the precision that detection is improved while epipole detection consistency is kept.
Description
Technical field
The invention belongs to living things feature recognition and the technical fields of information security, in particular to a kind of special based on image airspace
The fingerprint epipole accurate positioning method of sign.
Background technique
Biometrics identification technology is because biological characteristic (fingerprint, iris, face picture etc.) itself intrinsic characteristic makes it
Surmount and substitute the possibility that traditional identity recognizing technology becomes a reality, and in the specific application of some countries and regions
Field starts expanded and uses.Wherein fingerprint recognition in each item index of identification such as, generality, can adopt at uniqueness
Collection property, user's acceptance etc. have reached good balance, so that it has become a hot topic of research;In addition fingerprint collecting equipment and technology at
The advantages such as ripe, cheap and fingerprint template storage is low to space requirement, so that it occupies global biological characteristic market
The leading position of share.But still thering are some problems demands to solve, fingerprint singularity detection is one of them.According to Henry
Definition, singular point is divided into epipole --- the highest point of penetralia bending crestal line;And triangulation point --- three crestal lines are crossed into
Delta-shaped region central point.For existing fingerprint identification technology, most of inflection point detection methods are all based on block grade
The field of direction, therefore the precision of the singular point detected and consistency are all limited to the size of block, are embodied in:First, nothing
Singular point is accurately located to some pixel by method;Second, when two singular points are separated by close (in same piece),
The case where will appear missing inspection.Two main problems can be brought for subsequent identification:First, reducing on the basis of singular point
The efficiency of fingerprint image alignment schemes;Second, increasing the error rate of fingerprint classification method.
Inflection point detection method based on the Pixel-level field of direction is one of the approach to solve the above problems, and this method is not only
It is able to detect and is separated by close singular point pair, additionally it is possible to by singular point location to Pixel-level.But this method detects
Epipole still remain certain offset with the epipole that Henry is defined, and when nearby the higher hour offset of crestal line curvature is more for epipole
Seriously.The main reason for generating offset is that orientation estimate is inaccurate --- the low-pass filter introduced to denoise will be high
The field of direction of area of curvature is smooth together.The precision for wanting to further increase orientation estimate is highly difficult, it is intended to be retained
High curvature areas will certainly introduce more noises.Thus, we expect using fingerprint image spatial feature as auxiliary with essence
Determine position epipole.
The documents and materials having disclosed at present, there are no carry out epipole accurate positioning about based on fingerprint image spatial information (si)
Etc. related fields research.
Summary of the invention
The problem of being encountered in pushing large-scale practical application to the purpose of the present invention is to existing fingerprint identification technology,
Propose a kind of fingerprint epipole accurate positioning method based on image spatial feature.
The present invention specifically comprises the following steps:
Step 1, the Pixel-level field of direction O that fingerprint image is found out with gradient method;
Step 2 extracts each thick epipole position C with Poincare indexing method in field of direction Oi(cxi,cyi);
Step 3, the binary image I that fingerprint image is calculated with Short Time Fourier Transform method;
Step 4, by field of direction O, thick epipole position Ci(cxi,cyi) and binary image I input bifurcated polymerization extraction mould
Block extracts thick epipole position Ci(cxi,cyi) neighbouring all bifurcation fc and congruent point jh in a × b rectangular extent, and deposit
Storage is in matrix F JiIn;
Step 5, by matrix F JiDenoising module is input to remove matrix F JiAll pseudo- bifurcation, pseudo- congruent point, obtain
Matrix
Step 6, the matrix that will be obtained after denoisingIt is input to arc extraction module and obtains all arcs, and the arc that will be obtained
It is stored in matrix H u;
Step 7, matrix H u is input to pinpoint module obtain epipole positioning final result.
The polymerization extraction module includes following four submodules:
4-1. submodule 1 generates streakline width matrix Ei, streakline type matrix YSiWith streakline numbers matrix Zi;
4-1-1. calculates the argument principal value θ of thick epipole opening direction with existing algorithmi, specific as follows:
(formula 1)
Wherein, Oref(x, y) indicates that the reference direction field block of epipole is calculated having a size of 25 × 25 by formula 2;Oi
(x, y) is indicated with thick epipole Ci(cxi,cyi) centered on, on the field of direction O size of screenshot be 25 × 25 field of direction block;θref
For reference direction field block OrefThe argument principal value of the opening direction of (x, y);
(formula 2)
Wherein,For intermediate variable, Oref(x, y) is required reference direction field block;
4-1-2. is according to thick epipole position Ci(cxi,cyi) and corresponding argument principal value θi, make one on bianry image I
The a length of k pixel of item, the line segment L that width is single pixel1, so that thick epipole position Ci(cxi,cyi) it is located at line segment L1It is upper to lean on close over
Trisection point position;With line segment L1On each pixel be midpoint, make k item and line segment L1Vertically, and length is l
The line segment L of the single pixel wide of pixel2, to obtain the rectangle having a size of k × l;With the argument master of thick epipole opening direction
Value θiFor underface, every line segment L is successively recorded from left to right2The width of each fingerprint ridge passed through obtains streakline width
Matrix Ei;For record across the type of every streak line, 1 indicates that valley line, -1 indicate crestal line simultaneously, obtains streakline type matrix YSi;Point
The sum for not recording the streakline that every a line passes through obtains streakline numbers matrix Zi;The Ei、YSiSize be k × lfj,
Middle lfjFor ZiThe maximum value of middle element;ZiSize be k × 1;
It is edge compensation submodule that 4-2., which is calculated and obtained submodule 2, with streakline width matrix EiWith streakline type matrix YSi
As input, to edge compensation factor cw0、cw1Carry out assignment;
CYSi=YSi(j+1,1)×YSi(j, 1) (formula 3)
If CYSi> 0 then directly exports result cw0←0、cw1←0;Otherwise Δ E is calculatedi=Ei(j+1,1)-Ei(j, 1),
If Δ Ei>=0, then export result cw0←1、cw1← 0, otherwise export result cw0←0、cw1←1;
Wherein, CYSiWith Δ EiIt is intermediate variable, j is positive integer;
It is bifurcated type identification submodule that 4-3., which is calculated and obtained submodule 3,
(formula 4)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
With streakline width matrix Ei, streakline numbers matrix Zi, edge compensation factor cw0And cw1, bifurcated extract operator cz0(j,
W), the accumulated value a of skip over0And a1As input, to bifurcated type fc and its corresponding skip over ty1Carry out assignment;
4-3-1. to intermediate variable bc0And c0Carry out assignment, bc0←0、c0←cz0+Ei(j+1,w1+ 2) it then calculates intermediate
Variable g01And g02;
(formula 5)
(formula 6)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
If 4-3-2. g01≤g02Or w1+4+2bc0> Zi(j+1,1) then directly exports result fc ← 2bc0+2、ty1
←2bc0+ 2, jump out the submodule;Otherwise bc0←bc0After+1, step 4-3-1 is repeated;
It is that polymeric type differentiates submodule that 4-4., which is calculated and obtained submodule 4,
(formula 7)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
With streakline width matrix Ei, streakline numbers matrix Zi, edge compensation factor cw0And cw1, bifurcated extract operator cz1(j,
W), the accumulated value a of skip over0And a1As input, to polymeric type jh and corresponding skip over ty0Carry out assignment;
4-4-1. to intermediate variable bc1And c1Carry out assignment, bc1←0、c1←cz1+Ei(j,w0+ 2) intermediate become then is calculated
Measure g11And g12;
(formula 8)
(formula 9)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
If 4-4-2. g11≤g12Or w0+4+2bc1> Zi(j, 1) then directly exports result jh ← 2bc1+2、ty0←
2bc1+ 2, jump out the submodule;Otherwise bc1←bc1After+1, step 4-4-1 is repeated;
The extraction and storage of bifurcation fc and congruent point jh described in step 4 are specific as follows:
4-5. is by field of direction O, thick epipole position Ci(cxi,cyi) and binary image I be input to submodule 1 to find out Ei、
YSi、Zi;Then it is arranged one and EiThe identical null matrix FJ of sizei;And initial value, j ← 0 are assigned to positive integer j;
After 4-6.j ← j+1, by edge compensation submodule, E is calculatediCorresponding edge compensation factor cw0And cw1;And it is right
The accumulated value a of skip over0And a1, positive integer w assign initial value, a0←0、a1←0,w←1;
4-7. calculates w0=w+cw0+a0、w1=w+cw1+a1, then judged, if w0+1≤Zi(j, 1) and w1+1
≤Zi(j+1,1) then then calculates bifurcated with formula 4 and extracts operator cz0(j, w) calculates polymerization with formula 7 and extracts operator
cz1(j, w) goes to step 4-7-1:Otherwise step 4-6 is repeated, until completing to EiTraversal, to obtain corresponding matrix
FJi;
If 4-7-1. cz0(j, w) < 0 then finds out bifurcated type fc and its corresponding skip over by submodule 3
ty1, bifurcated is then saved into FJi(j+1,w1+ fc/2) ← fc, and update cumulative skip over a1Value, a1←ty1+a1, go to
Step 4-7-2;Otherwise step 4-7-2 is directly gone to;
If 4-7-2. cz1(j, w) < 0 then finds out polymeric type jh and its corresponding skip over by submodule 4
ty0, polymerization is then saved into FJi(j+1,w0) ← jh, and update cumulative skip over a0Value, a0←ty0+a0;
4-7-3.a0←0、a1← 0, w ← w+1 repeats step 4-7;
In FJiIn, FJi(x, y)=2 indicates EiMiddle xth row y column are a bifurcated, FJi(x, y)=4 indicates to correspond to
Position is four bifurcateds, FJi(x, y)=6 indicates that corresponding position is six bifurcateds, FJi(x, y)=8 indicates that corresponding position is a eight
Bifurcated,;Otherwise FJ is then used in polymerizationi(x, y)=- 2, FJi(x, y)=- 4, FJi(x, y)=- 6, FJi(x, y)=- 8 is indicated.
Denoising module described in step 5 includes following two submodules:
It is edge compensation vector calculation module that 5-1., which is calculated and obtained submodule 5, with streakline width matrix EiWith streakline type
Matrix YSiAs input, assignment is carried out to edge compensation vector B;
Null vector B of one length of 5-1-1. application for k, j ← 1, then;
5-1-2. calls submodule 2 to calculate cw0、cw1;Calculate B (1, j+1)=B (1, j)+cw1-cw0;
5-1-3. terminates operation if j >=k-1;Otherwise step 5-1-2 is repeated after j ← j+1;
5-2., which is calculated, obtains submodule 6 i.e. with streakline differentiation submodule, with k02、l02、k2、l2、FJiAs input, to twx2
Carry out assignment;
5-2-1. calculates wz02=l2+B(1,k2)-FJi(k2,l2)/2、wz12=l02+B(1,k02)-FJi(k02,l02)/2;
Wherein B is the vector that submodule 5 is calculated;
5-2-2. is with FJi(k02It+1,1) is head, with FJi(k2,l2It -1) is tail, with sequence time from left to right, from top to bottom
Go through FJi, and use FJi(k12,l12) indicate wherein each element;If FJi(k12,l12) ≠ 0, then calculate wz22=l12+B(1,
k12)-FJi(k12,l12)/2, then, if wz22≤wz12, then wz is updated12Value, wz12←wz12+FJi(k12,l12);
After 5-2-3. completes traversal, p is calculated2=| wz02-wz12|、t2=max (2, | FJi(k02,l02)/2), if p2<
t2, then twx2← 1, otherwise twx2←0;
Denoising module described in step 5 realizes that steps are as follows:
5-3. is with A, FJiAs input, wherein [- 8-6-4-2] A=, output isBy FJiIt is assigned toj2←
1, a2←A(1,j2);
5-4. traversalWithIt indicatesIn any one element, if there is noThen j2←j2A is returned after+12←A(1,j2), and enter and traverse next time;Otherwise, km2←max(k2-t1,
0) it, traversesMiddle rower is located at section [km2,k2] element, and withIndicate it, if
Then by k02、l02、k2、l2、FJiIt is input to submodule 6 and finds out twx;
If the twx==1 that 5-5. is found out, and meet simultaneously, if otherwise not meeting simultaneously
Then
After 5-6. has been traversed, j2←j2+ 1, a2←A(1,j2), step 5-4 and 5-5 are repeated, until four elements are all complete in A
At primary rightTraversal, after having traversedThe final result as denoised.
Epipole pinpoint module described in step 6 includes following two submodules:
6-1., which is calculated, obtains submodule 7 i.e. with streakline differentiation submodule II, with streakline width matrix EiWith streakline type square
Battle array YSiAs input, assignment is carried out to edge compensation vector B;
6-2., which is calculated, obtains submodule 7 i.e. with streakline differentiation submodule II, with k02、l02、k2、l2、FJiIt is right as input
twx3Carry out assignment;
6-2-1. calculates wz02=l2+B(1,k2)-FJi(k2,l2)/2、、wz12=l02+B(1,k02)-FJi(k02,l02)/2;
Wherein B is the vector that submodule 5 is calculated;
6-2-2. is with FJi(k02It+1,1) is head, with FJi(k2,l2It -1) is tail, with sequence time from left to right, from top to bottom
Go through FJi, and use FJi(k12,l12) indicate wherein each element;If FJi(k12,l12) ≠ 0, then calculate wz22=l12+B(1,
k12)-FJi(k12,l12)/2, then, if wz22≤wz12, then wz is updated12Value, wz12←wz12+FJi(k12,l12);
After 6-2-3. completes traversal, p is calculated3=wz12-wz02、t3If=2 p3< t3, then twx3← 1, otherwise twx3←
0。
As shown in figure 4, submodule 8 described in step 7 is that arc extraction module is specific as follows:
7-1. is with FJi、YSiAs input, assignment is carried out to matrix H u;
7-1-1.k23← 0, the null matrix Hu that application is one 1 × 1;
7-1-2. traversalYSi, withYSi(k03,l03) respectively indicateYSiIn element;Such as
FruitAnd YSi(k03,l03)==-1, then k23←k23+ 1, then traverseYSiMiddle row coordinate is located at section
[km3,k03] (k herem3=min (k03+t03,kfj), t03It is determined by the resolution ratio of fingerprint image, for the image of 500dpi,
Value range is t03∈ [15,25]) element, and withYSi(k13,l13) it is respectively indicated, ifAnd YSi(k13,l13)==1 then calculates twx by submodule 73;If twx3==1, then by array
[k03 l03] it is assigned to Hu (k23, 1), array [k13 l13] it is assigned to Hu (k23,2)。
Pinpoint module described in step 8 is specific as follows:
8-1. is by FJi、YSiAs input, with thick epipole CiCorresponding accurate positioning resultAs output;
8-2. Ergodic Matrices Hu first, all elements assignment of the row where wherein the last one nonzero element is given to
Amount
8-3. willValue be assigned to array [k33 l33], traversalYSiPositioned at section [k33+1,km4] row,
And withYSi(k43,l43) indicate wherein element;If there isAnd YSi(k43,l43)==
1, then by array [k at this time43 l43] be assigned toThen it jumps out the traversal and then terminates whole flow process;Otherwise by array
[k33 l33] be assigned toThen terminate whole flow process.
The present invention has the beneficial effect that:
Fingerprint epipole accurate positioning method the present invention is based on image spatial information (si) includes orientation estimate module, singular point
Detection module, image binaryzation module, bifurcated, polymerization extraction module denoise module, epipole pinpoint module.
Since the present invention is based on image spatial information (si), for epipole detection, spatial information (si) is for detection accuracy
Good support is provided, such precision is difficult to realize in the epipole detection method for being based only upon the field of direction.
Since the present invention is accurately positioned in unit in epipole, use is all linear operation, is guaranteeing positioning accuracy
Lower computing cost is also maintained simultaneously.
Since the present invention introduces easy and suitable denoising method, thus whole system pair when detecting bifurcated, polymerization
Binaryzation has no particular/special requirement, improves this method to the friendliness of fingerprint binarization method.
Detailed description of the invention
Fig. 1 is the schematic diagram of the argument principal value of the field of direction block and direction field block opening around thick epipole of the invention.
Fig. 2 is the schematic diagram of " bifurcated " of the invention.
Fig. 3 is the schematic diagram of " polymerization " of the invention.
Fig. 4 is the schematic diagram of " arc " of the invention.
Fig. 5 is matrix E of the present inventioni、YSi、ZiThe schematic diagram of generating mode.
Fig. 6 is " skip over " schematic diagram of the invention.
Specific embodiment
Below with reference to embodiment, the present invention is further described.
Epipole accurate positioning method based on image spatial feature, specifically comprises the following steps:
Step 1, the Pixel-level field of direction O that fingerprint image is found out with gradient method;
Step 2 extracts each thick epipole position with Poincare indexing method (poincare index) in field of direction O
Ci(cxi,cyi);
Step 3, the binary image I that fingerprint image is calculated with Short Time Fourier Transform method (STFT);
Step 4, by field of direction O, thick epipole position Ci(cxi,cyi) and binary image I input bifurcated polymerization extraction mould
Block extracts thick epipole position Ci(cxi,cyi) neighbouring all bifurcation fc and congruent point jh in a × b rectangular extent, and deposit
Storage is in matrix F JiIn;
Step 5, by matrix F JiDenoising module is input to remove matrix F JiAll pseudo- bifurcation, pseudo- congruent point, obtain
Matrix
Step 6, the matrix that will be obtained after denoisingIt is input to arc extraction module and obtains all arcs, and the arc that will be obtained
It is stored in matrix H u;
Step 7, matrix H u is input to pinpoint module obtain epipole positioning final result.
Polymerization extraction module described in step 4 includes following four submodules:
4-1. submodule 1 generates streakline width matrix Ei, streakline type matrix YSiWith streakline numbers matrix Zi。
4-1-1. calculates the argument principal value θ of thick epipole opening direction with existing algorithmi, specific as follows:
(formula 1)
As shown in Figure 1, wherein Oref(x, y) indicates that the reference direction field block of epipole is counted having a size of 25 × 25 by formula 2
It obtains;Oi(x, y) is indicated with thick epipole Ci(cxi,cyi) centered on, on the field of direction O size of screenshot be 25 × 25 side
To field block;θrefFor reference direction field block OrefThe argument principal value of the opening direction of (x, y).
(formula 2)
Wherein,For intermediate variable, Oref(x, y) is required reference direction field block.
4-1-2. is according to thick epipole position Ci(cxi,cyi) and corresponding argument principal value θi, make one on bianry image I
The a length of k pixel of item, the line segment L that width is single pixel1(as shown in the grey line segment in the area Tu5Zhong a), so that thick epipole position Ci(cxi,
cyi) it is located at line segment L1It is upper to lean on close over (with the argument principal value θ of thick epipole opening directioniDirection be lower section) trisection point
Position;With line segment L1On each pixel be midpoint, make k item and line segment L1Vertically, and length be l pixel single picture
The wide line segment L of element2(as shown in the grey line segment in the area Tu5Zhong b), to obtain rectangle (k, a l here having a size of k × l
Value determined by the resolution ratio of fingerprint image, when image be 500dpi when, k ∈ [50,70], l ∈ [40,55]);With thick epipole
The argument principal value θ of opening directioniFor underface, every line segment L is successively recorded from left to right2The width of each fingerprint ridge passed through
Degree, obtains streakline width matrix Ei(such as the area Tu5Zhong c is shown), while the type across every streak line is recorded, 1 indicates paddy
Line, -1 indicate crestal line, obtain streakline type matrix YSi(such as the area Tu5Zhong d is shown) records the streakline that every a line passes through respectively
Sum obtain streakline numbers matrix Zi(such as the area Tu5Zhong e is shown), here Ei、YSiSize be k × lfj(lfjFor ZiIn
The maximum value of element);ZiSize be k × 1.
It is edge compensation submodule that 4-2., which is calculated and obtained submodule 2, with streakline width matrix EiWith streakline type matrix YSi
As input, to edge compensation factor cw0、cw1Carry out assignment.
CYSi=YSi(j+1,1)×YSi(j, 1) (formula 3)
If CYSi> 0 then directly exports result cw0←0、cw1←0;Otherwise Δ E is calculatedi=Ei(j+1,1)-Ei(j, 1),
If Δ Ei>=0, then export result cw0←1、cw1← 0, otherwise export result cw0←0、cw1←1。
Wherein, CYSiWith Δ EiIt is intermediate variable, j is positive integer;
It is bifurcated type identification submodule that 4-3., which is calculated and obtained submodule 3,
(formula 4)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer.
With streakline width matrix Ei, streakline numbers matrix Zi, edge compensation factor cw0And cw1, bifurcated extract operator cz0(j,
W), the accumulated value a of skip over0And a1As input, to bifurcated type fc and its corresponding skip over ty1Carry out assignment.
(as shown in fig. 6, box B is the corresponding skip over of dimerization, box A is the corresponding skip over of bifurcated, it
Will additionally increase cz next time0Summation section length;The value of the corresponding skip over of four bifurcateds is 4, and the rest may be inferred).
4-3-1. to intermediate variable bc0And c0Carry out assignment, bc0←0、c0←cz0+Ei(j+1,w1+ 2) it then calculates intermediate
Variable g01And g02;
(formula 5)
(formula 6)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer.
If 4-3-2. g01≤g02Or w1+4+2bc0> Zi(j+1,1) then directly exports result fc ← 2bc0+2、ty1
←2bc0+ 2, jump out the submodule;Otherwise bc0←bc0After+1, step 4-3-1 is repeated;
It is that polymeric type differentiates submodule that 4-4., which is calculated and obtained submodule 4,
(formula 7)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer.
With streakline width matrix Ei, streakline numbers matrix Zi, edge compensation factor cw0And cw1, bifurcated extract operator cz1(j,
W), the accumulated value a of skip over0And a1As input, to polymeric type jh and corresponding skip over ty0Carry out assignment.
4-4-1. to intermediate variable bc1And c1Carry out assignment, bc1←0、c1←cz1+Ei(j,w0+ 2) intermediate become then is calculated
Measure g11And g12;
(formula 8)
(formula 9)
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer.
If 4-4-2. g11≤g12Or w0+4+2bc1> Zi(j, 1) then directly exports result jh ← 2bc1+2、ty0←
2bc1+ 2, jump out the submodule;Otherwise bc1←bc1After+1, step 4-4-1 is repeated.
As shown in Figures 2 and 3, the extraction and storage of bifurcation fc and congruent point jh described in step 4 are specific as follows:
4-5. is by field of direction O, thick epipole position Ci(cxi,cyi) and binary image I be input to submodule 1 to find out Ei、
YSi、Zi;Then it is arranged one and EiThe identical null matrix FJ of sizei;And initial value, j ← 0 are assigned to positive integer j;
After 4-6.j ← j+1, by edge compensation submodule (i.e. step 4-2), E is calculatediCorresponding edge compensation factor cw0
And cw1;And to the accumulated value a of skip over0And a1, positive integer w assign initial value, a0←0、a1←0,w←1;
4-7. calculates w0=w+cw0+a0、w1=w+cw1+a1, then judged, if w0+1≤Zi(j, 1) and w1+1
≤Zi(j+1,1) then then calculates bifurcated with (formula 4) and extracts operator cz0(j, w) calculates polymerization with (formula 7) and extracts
Operator cz1(j, w) goes to step 4-7-1:Otherwise step 4-6 is repeated, until completing to EiTraversal, to obtain corresponding square
Battle array FJi。
If 4-7-1. cz0(j, w) < 0 then finds out bifurcated type fc and its correspondence by submodule 3 (step 4-3)
Skip over ty1, bifurcated is then saved into FJi(j+1,w1+ fc/2) ← fc, and update cumulative skip over a1Value, a1←
ty1+a1, go to step 4-7-2;Otherwise step 4-7-2 is directly gone to;
If 4-7-2. cz1(j, w) < 0 then finds out polymeric type jh and its correspondence by submodule 4 (step 4-4)
Skip over ty0, polymerization is then saved into FJi(j+1,w0) ← jh, and update cumulative skip over a0Value, a0←ty0+
a0;
4-7-3.a0←0、a1← 0, w ← w+1 repeats step 4-7;
It should be strongly noted that in FJiIn, FJi(x, y)=2 indicates EiMiddle xth row y column are a bifurcated, FJi
(x, y)=4 indicates that corresponding position is four bifurcateds, FJi(x, y)=6 indicates that corresponding position is six bifurcateds, FJi(x, y)=8 table
Show that corresponding position is eight bifurcateds;Otherwise FJ is then used in polymerizationi(x, y)=- 2, FJi(x, y)=- 4, FJi(x, y)=- 6, FJi(x,
Y) it=- 8 indicates.
Denoising module described in step 5 includes following two submodules:
It is edge compensation vector calculation module that 5-1., which is calculated and obtained submodule 5, with streakline width matrix EiWith streakline type
Matrix YSiAs input, assignment is carried out to edge compensation vector B.
Null vector B of one length of 5-1-1. application for k, j ← 1, then;
5-1-2. calls submodule 2 to calculate cw0、cw1;Calculate B (1, j+1)=B (1, j)+cw1-cw0;
5-1-3. terminates operation if j >=k-1;Otherwise step 5-1-2 is repeated after j ← j+1;
5-2., which is calculated, obtains submodule 6 i.e. with streakline differentiation submodule, with k02、l02、k2、l2、FJiAs input, to twx2
Carry out assignment.
5-2-1. calculates wz02=l2+B(1,k2)-FJi(k2,l2)/2、wz12=l02+B(1,k02)-FJi(k02,l02)/2;
Wherein B is the vector that submodule 5 is calculated;
5-2-2. is with FJi(k02It+1,1) is head, with FJi(k2,l2It -1) is tail, with sequence time from left to right, from top to bottom
Go through FJi, and use FJi(k12,l12) indicate wherein each element;If FJi(k12,l12) ≠ 0, then calculate wz22=l12+B(1,
k12)-FJi(k12,l12)/2, then, if wz22≤wz12, then wz is updated12Value, wz12←wz12+FJi(k12,l12)。
After 5-2-3. completes traversal, p is calculated2=| wz02-wz12|、t2=max (2, | FJi(k02,l02)/2), if p2<
t2, then twx2← 1, otherwise twx2←0;
Denoising module described in step 5 realizes that steps are as follows:
5-3. is with A, FJiAs input, wherein [- 8-6-4-2] A=, output isBy FJiIt is assigned toj2
← 1, a2←A(1,j2);
5-4. traversalWithIt indicatesIn any one element, if there is noThen j2←j2A is returned after+12←A(1,j2), and enter and traverse next time;Otherwise, km2←max(k2-t1,
0) (t here1It is determined by the resolution ratio of fingerprint image, under the image of 500dpi, value range is t1∈ [7,18]), traversalMiddle rower is located at section [km2,k2] element, and withIndicate it, ifThen will
k02、l02、k2、l2、FJiIt is input to submodule 6 and finds out twx;
If the twx==1 that 5-5. is found out, and meet simultaneously, if otherwise not meeting simultaneously
Then
After 5-6. has been traversed, j2←j2+ 1, a2←A(1,j2), step 5-4 and 5-5 are repeated, until four elements are all complete in A
At primary rightTraversal, after having traversedThe final result as denoised.
Epipole pinpoint module described in step 6 includes following two submodules:
6-1., which is calculated, obtains submodule 7 i.e. with streakline differentiation submodule II, with streakline width matrix EiWith streakline type square
Battle array YSiAs input, assignment is carried out to edge compensation vector B.
6-2., which is calculated, obtains submodule 7 i.e. with streakline differentiation submodule II, with k02、l02、k2、l2、FJiIt is right as input
twx3Carry out assignment.
6-2-1. calculates wz02=l2+B(1,k2)-FJi(k2,l2)/2、、wz12=l02+B(1,k02)-FJi(k02,l02)/2;
Wherein B is the vector that submodule 5 is calculated;
6-2-2. is with FJi(k02It+1,1) is head, with FJi(k2,l2It -1) is tail, with sequence time from left to right, from top to bottom
Go through FJi, and use FJi(k12,l12) indicate wherein each element;If FJi(k12,l12) ≠ 0, then calculate wz22=l12+B(1,
k12)-FJi(k12,l12)/2, then, if wz22≤wz12, then wz is updated12Value, wz12←wz12+FJi(k12,l12)。
After 6-2-3. completes traversal, p is calculated3=wz12-wz02、t3If=2 p3< t3, then twx3← 1, otherwise twx3←
0。
Submodule 8 described in step 7 is that arc extraction module is specific as follows:
7-1. is with FJi、YSiAs input, assignment is carried out to matrix H u.
7-1-1.k23← 0, the null matrix Hu that application is one 1 × 1;
7-1-2. traversalYSi, withYSi(k03,l03) respectively indicateYSiIn element;IfAnd YSi(k03,l03)==-1, then k23←k23+ 1, then traverseYSiMiddle row coordinate is located at section
[km3,k03] (k herem3=min (k03+t03,kfj), t03It is determined by the resolution ratio of fingerprint image, for the image of 500dpi,
Value range is t03∈ [15,25]) element, and withYSi(k13,l13) it is respectively indicated, ifAnd YSi(k13,l13)==1 then calculates twx by submodule 73;If twx3==1, then by array
[k03 l03] it is assigned to Hu (k23, 1), array [k13 l13] it is assigned to Hu (k23, 2) and (here, primary traversal is likely to be obtained multiple full
[the k of sufficient condition13 l13], they are assigned to Hu (k respectively23,3)、Hu(k23, 4), and so on;The size of Hu is with assignment
It is how many and change, and zero) initial value of each element is.
Pinpoint module described in step 8 is specific as follows:
8-1. is by FJi、YSiAs input, with thick epipole CiCorresponding accurate positioning result CiAs output.
8-2. Ergodic Matrices Hu first, all elements assignment of the row where wherein the last one nonzero element is given to
Amount
8-3. willValue be assigned to array [k33 l33], traversalYSiPositioned at section [k33+1,km4] (here
Km4=min (k33+t13, k), t13It is determined by the resolution ratio of fingerprint image, for the image of 500dpi, t13∈ [10,18])
Row, and withYSi(k43,l43) indicate wherein element;If there isAnd YSi(k43,l43)
==1, then by array [k at this time43 l43] be assigned toThen it jumps out the traversal and then terminates whole flow process;Otherwise it will count
Group [k33 l33] be assigned toThen terminate whole flow process.
Claims (5)
1. the fingerprint epipole accurate positioning method based on image spatial feature, it is characterised in that include the following steps:
Step 1, the Pixel-level field of direction O that fingerprint image is found out with gradient method;
Step 2 extracts each thick epipole position C with Poincare indexing method in field of direction Oi(cxi,cyi);
Step 3, the binary image I that fingerprint image is calculated with Short Time Fourier Transform method;
Step 4, by field of direction O, thick epipole position Ci(cxi,cyi) and binary image I input bifurcated polymerization extraction module, it mentions
Take out thick epipole position Ci(cxi,cyi) neighbouring all bifurcation fc and congruent point jh in a × b rectangular extent, and it is stored in square
Battle array FJiIn;
Step 5, by matrix F JiDenoising module is input to remove matrix F JiAll pseudo- bifurcation, pseudo- congruent point, obtain matrix
Step 6, the matrix that will be obtained after denoisingIt is input to arc extraction module and obtains all arcs, and obtained arc is stored
In matrix H u, matrix is utilizedDifferentiate with line, calculating parameter twx3
Step 7, matrix H u is input to pinpoint module obtain epipole positioning final result;
It includes following four submodules that bifurcated described in step 4, which polymerize extraction module,:
4-1. submodule 1 generates streakline width matrix Ei, streakline type matrix YSiWith streakline numbers matrix Zi;
4-1-1. calculates the argument principal value θ of thick epipole opening direction with existing algorithmi, specific as follows:
Wherein, Oref(x, y) indicates that the reference direction field block of epipole is calculated having a size of 25 × 25 by formula 2;Oi(x, y) table
Show with thick epipole Ci(cxi,cyi) centered on, on the field of direction O size of screenshot be 25 × 25 field of direction block;θrefFor reference
Field of direction block OrefThe argument principal value of the opening direction of (x, y);
Wherein,For intermediate variable, Oref(x, y) is required reference direction field block;
4-1-2. is according to thick epipole position Ci(cxi,cyi) and corresponding argument principal value θi, one, work is a length of on bianry image I
K pixel, the line segment L that width is single pixel1, so that thick epipole position Ci(cxi,cyi) it is located at line segment L1It is upper to lean on the third of close over
The position of branch;With line segment L1On each pixel be midpoint, make k item and line segment L1Vertically, and length is l pixel
The line segment L of single pixel wide2, to obtain the rectangle having a size of k × l;With the argument principal value θ of thick epipole opening directioniIt is positive
Lower section successively records every line segment L from left to right2The width of each fingerprint ridge passed through obtains streakline width matrix Ei;Together
Shi Jilu passes through the type of every streak line, and 1 indicates that valley line, -1 indicate crestal line, obtains streakline type matrix YSi;Record is every respectively
The sum for the streakline that a line passes through obtains streakline numbers matrix Zi;The Ei、YSiSize be k × lfj, wherein lfjFor Zi
The maximum value of middle element;ZiSize be k × 1;
It is edge compensation submodule that 4-2., which is calculated and obtained submodule 2, with streakline width matrix EiWith streakline type matrix YSiAs
Input, to edge compensation factor cw0、cw1Carry out assignment;
CYSi=YSi(j+1,1)×YSi(j, 1) (formula 3)
If CYSi> 0 then directly exports result cw0←0、cw1←0;Otherwise Δ E is calculatedi=Ei(j+1,1)-Ei(j, 1), if
ΔEi>=0, then export result cw0←1、cw1← 0, otherwise export result cw0←0、cw1←1;
Wherein, CYSiWith Δ EiIt is intermediate variable, j is positive integer;
It is bifurcated type identification submodule that 4-3., which is calculated and obtained submodule 3,
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
With streakline width matrix Ei, streakline numbers matrix Zi, edge compensation factor cw0And cw1, bifurcated extract operator cz0(j,w)、
The accumulated value a of skip over0And a1As input, to bifurcated type fc and its corresponding skip over ty1Carry out assignment;
4-3-1. to intermediate variable bc0And c0Carry out assignment, bc0←0、c0←cz0+Ei(j+1,w1+ 2) intermediate variable is then calculated
g01And g02;
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
If 4-3-2. g01≤g02Or w1+4+2bc0> Zi(j+1,1) then directly exports result fc ← 2bc0+2、ty1←2bc0
+ 2, jump out the submodule;Otherwise bc0←bc0After+1, step 4-3-1 is repeated;
It is that polymeric type differentiates submodule that 4-4., which is calculated and obtained submodule 4,
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
With streakline width matrix Ei, streakline numbers matrix Zi, edge compensation factor cw0And cw1, bifurcated extract operator cz1(j,w)、
The accumulated value a of skip over0And a1As input, to polymeric type jh and corresponding skip over ty0Carry out assignment;
4-4-1. to intermediate variable bc1And c1Carry out assignment, bc1←0、c1←cz1+Ei(j,w0+ 2) intermediate variable g is then calculated11
And g12;
Wherein, j is positive integer, w0=w+cw0+a0、w1=w+cw1+a1, w is positive integer;
If 4-4-2. g11≤g12Or w0+4+2bc1> Zi(j, 1) then directly exports result jh ← 2bc1+2、ty0←2bc1+
2, jump out the submodule;Otherwise bc1←bc1After+1, step 4-4-1 is repeated;
The extraction and storage of bifurcation fc and congruent point jh described in step 4 are specific as follows:
4-5. is by field of direction O, thick epipole position Ci(cxi,cyi) and binary image I be input to submodule 1 to find out Ei、YSi、
Zi;Then it is arranged one and EiThe identical null matrix FJ of sizei;And initial value, j ← 0 are assigned to positive integer j;
After 4-6.j ← j+1, by edge compensation submodule, E is calculatediCorresponding edge compensation factor cw0And cw1;And to jump
The accumulated value a of the factor0And a1, positive integer w assign initial value, a0←0、a1←0,w←1;
4-7. calculates w0=w+cw0+a0、w1=w+cw1+a1, then judged, if w0+1≤Zi(j, 1) and w1+1≤Zi
(j+1,1) then then calculates bifurcated with formula 4 and extracts operator cz0(j, w) calculates polymerization with formula 7 and extracts operator cz1
(j, w) goes to step 4-7-1:Otherwise step 4-6 is repeated, until completing to EiTraversal, to obtain corresponding matrix F Ji:
If 4-7-1. cz0(j, w) < 0 then finds out bifurcated type fc and its corresponding skip over ty by submodule 31,
Then bifurcated is saved into FJi(j+1,w1+ fc/2) ← fc, and update cumulative skip over a1Value, a1←ty1+a1, go to step
4-7-2;Otherwise step 4-7-2 is directly gone to;
If 4-7-2. cz1(j, w) < 0 then finds out polymeric type jh and its corresponding skip over ty by submodule 40,
Then polymerization is saved into FJi(j+1,w0) ← jh, and update cumulative skip over a0Value, a0←ty0+a0;
4-7-3.a0←0、a1← 0, w ← w+1 repeats step 4-7;
In FJiIn, FJi(x, y)=2 indicates EiMiddle xth row y column are a bifurcated, FJi(x, y)=4 indicates that corresponding position is
A four bifurcated, FJi(x, y)=6 indicates that corresponding position is six bifurcateds, FJi(x, y)=8 indicates that corresponding position is eight bifurcateds;
Otherwise FJ is then used in polymerizationi(x, y)=- 2, FJi(x, y)=- 4, FJi(x, y)=- 6, FJi(x, y)=- 8 is indicated.
2. the fingerprint epipole accurate positioning method based on image spatial feature as described in claim 1, it is characterised in that step 5
The denoising module includes following two submodules:
It is edge compensation vector calculation module that 5-1., which is calculated and obtained submodule 5, with streakline width matrix EiWith streakline type matrix
YSiAs input, assignment is carried out to edge compensation vector B;
Null vector B of one length of 5-1-1. application for k, j ← 1, then;
5-1-2. calls submodule 2 to calculate cw0、cw1;Calculate B (1, j+1)=B (1, j)+cw1-cw0;
5-1-3. terminates operation if j >=k-1;Otherwise step 5-1-2 is repeated after j ← j+1;
5-2., which is calculated, obtains submodule 6 i.e. with streakline differentiation submodule, with k02、l02、k2、l2、FJiAs input, to twx2It carries out
Assignment;
5-2-1. calculates wz02=l2+B(1,k2)-FJi(k2,l2)/2、wz12=l02+B(1,k02)-FJi(k02,l02)/2;Wherein B
It is the vector that submodule 5 is calculated;
5-2-2. is with FJi(k02It+1,1) is head, with FJi(k2,l2It -1) is tail, with order traversal from left to right, from top to bottom
FJi, and use FJi(k12,l12) indicate wherein each element;If FJi(k12,l12) ≠ 0, then calculate wz22=l12+B(1,
k12)-FJi(k12,l12)/2, then, if wz22≤wz12, then wz is updated12Value, wz12←wz12+FJi(k12,l12);
After 5-2-3. completes traversal, p is calculated2=| wz02-wz12|、t2=max (2, | FJi(k02,l02)/2 |), if p2< t2, then
twx2← 1, otherwise twx2←0;
Denoising module described in step 5 realizes that steps are as follows:
5-3. with A, FJiAs input, wherein [- 8-6-4-2] A=, output isBy FJiIt is assigned toj2← 1, a2
←A(1,j2);
5-4. traversalWithIt indicatesIn any one element, if there is noThen j2←j2A is returned after+12←A(1,j2), and enter and traverse next time;Otherwise, km2←max(k2-
t1, 0), traversalMiddle rower is located at section [km2,k2] element, and withIndicate it, ifThen by k02、l02、k2、l2、FJiIt is input to submodule 6 and finds out twx2;
If the twx that 5-5. is found out2==1, and meet simultaneouslyThenIf otherwise not meeting simultaneously
Then
After 5-6. has been traversed, j2←j2+ 1, a2←A(1,j2), step 5-4 and 5-5 are repeated, until four elements all complete one in A
It is secondary rightTraversal, after having traversedThe final result as denoised.
3. the fingerprint epipole accurate positioning method based on image spatial feature as claimed in claim 2, it is characterised in that step 6
The arc extraction module is specific as follows:
6-1. is with FJi、YSiAs input, assignment is carried out to matrix H u;
6-1-1.k23← 0, the null matrix Hu that application is one 1 × 1;
6-1-2. traversalYSi, withYSi(k03,l03) respectively indicateYSiIn element;IfAnd YSi(k03,l03)==-1, then k23←k23+ 1, then traverseYSiMiddle row coordinate is located at section
[km3,k03] (k herem3=min (k03+t03,kfj), t03It is determined by the resolution ratio of fingerprint image, for the image of 500dpi,
Value range is t03∈ [15,25]) element, and withYSi(k13,l13) it is respectively indicated, ifAnd YSi(k13,l13)==1, then calculate twx3;If twx3==1, then by array [k03 l03] assign
It is worth and gives Hu (k23, 1), array [k13 l13] it is assigned to Hu (k23,2)。
4. the fingerprint epipole accurate positioning method based on image spatial feature as claimed in claim 3, it is characterised in that step 7
The pinpoint module is specific as follows:
It is edge compensation vector calculation module that 7-1., which is calculated and obtained submodule 7, with streakline width matrix EiWith streakline type matrix
YSiAs input, assignment is carried out to edge compensation vector B;
7-2., which is calculated, obtains submodule 8 i.e. with streakline differentiation submodule II, with k02、l02、k2、l2、FJiAs input, to twx3Into
Row assignment;
7-2-1. calculates wz02=l2+B(1,k2)-FJi(k2,l2)/2、\ wz12=l02+B(1,k02)-FJi(k02,l02)/2;Its
Middle B is the vector that submodule 7 is calculated;
7-2-2. is with FJi(k02It+1,1) is head, with FJi(k2,l2It -1) is tail, with order traversal from left to right, from top to bottom
FJi, and use FJi(k12,l12) indicate wherein each element;If FJi(k12,l12) ≠ 0, then calculate wz22=l12+B(1,
k12)-FJi(k12,l12)/2, then, if wz22≤wz12, then wz is updated12Value, wz12←wz12+FJi(k12,l12);
After 7-2-3. completes traversal, p is calculated3=wz12-wz02、t3If=2 p3< t3, then twx3← 1, otherwise twx3←0。
5. the fingerprint epipole accurate positioning method based on image spatial feature as claimed in claim 4, it is characterised in that described
Pinpoint module be implemented as follows:
8-1. is by FJi、YSiAs input, with thick epipole CiCorresponding accurate positioning resultAs output;
8-2. Ergodic Matrices Hu first, all elements of the row where wherein the last one nonzero element are assigned to vector
8-3. willValue be assigned to array [k33 l33], traversalYSiPositioned at section [k33+1, km4] row, and
WithYSi(k43,l43) indicate wherein element;If there isAnd YSi(k43,l43)==
1, then by array [k at this time43 l43] be assigned toThen it jumps out the traversal and then terminates whole flow process;Otherwise by array
[k33 l33] be assigned toThen terminate whole flow process.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510196807.1A CN104866815B (en) | 2015-04-23 | 2015-04-23 | Fingerprint epipole accurate positioning method based on image spatial feature |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510196807.1A CN104866815B (en) | 2015-04-23 | 2015-04-23 | Fingerprint epipole accurate positioning method based on image spatial feature |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104866815A CN104866815A (en) | 2015-08-26 |
CN104866815B true CN104866815B (en) | 2018-11-30 |
Family
ID=53912636
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510196807.1A Active CN104866815B (en) | 2015-04-23 | 2015-04-23 | Fingerprint epipole accurate positioning method based on image spatial feature |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104866815B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105389563B (en) * | 2015-11-16 | 2018-08-10 | 杭州电子科技大学 | Fingerprint puppet epipole detection method based on field of direction global information |
US9530042B1 (en) | 2016-06-13 | 2016-12-27 | King Saud University | Method for fingerprint classification |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1588424A (en) * | 2004-07-02 | 2005-03-02 | 清华大学 | Finger print identifying method based on broken fingerprint detection |
CN1595428A (en) * | 2004-07-15 | 2005-03-16 | 清华大学 | Fingerprint identification method based on density chart model |
CN1595425A (en) * | 2004-07-13 | 2005-03-16 | 清华大学 | Method for identifying multi-characteristic of fingerprint |
CN101303729A (en) * | 2008-07-01 | 2008-11-12 | 山东大学 | Novel method for detecting fingerprint singularity |
CN103500334A (en) * | 2013-09-17 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Method and device for acquiring direction fields of overlapped fingerprints |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9070002B2 (en) * | 2011-10-18 | 2015-06-30 | King Fahd University Of Petroleum And Minerals | Method of performing fingerprint matching |
-
2015
- 2015-04-23 CN CN201510196807.1A patent/CN104866815B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1588424A (en) * | 2004-07-02 | 2005-03-02 | 清华大学 | Finger print identifying method based on broken fingerprint detection |
CN1595425A (en) * | 2004-07-13 | 2005-03-16 | 清华大学 | Method for identifying multi-characteristic of fingerprint |
CN1595428A (en) * | 2004-07-15 | 2005-03-16 | 清华大学 | Fingerprint identification method based on density chart model |
CN101303729A (en) * | 2008-07-01 | 2008-11-12 | 山东大学 | Novel method for detecting fingerprint singularity |
CN103500334A (en) * | 2013-09-17 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Method and device for acquiring direction fields of overlapped fingerprints |
Non-Patent Citations (3)
Title |
---|
Fingerprint Singular Point Detection Based on Modified Poincare Index Method;Gabriel Babatunde Iwasokun 等;《International Journal of Signal Processing, Image Processing and Pattern Recognition》;20141231;第7卷(第5期);第269-272页 * |
指纹奇异点精确定位新方法;张祖泷 等;《计算机工程与应用》;20121111;第48卷(第32期);第203页摘要,第201页-204页第1部分引言,第204-206页第4部分奇异点检测方法 * |
指纹识别系统中的关键算法研究;刘波;《中国优秀博硕士学位论文全文数据库(硕士)·信息科技辑》;20060615(第06期);第12-14页第2.3-2.4节,第38-48页第4章 * |
Also Published As
Publication number | Publication date |
---|---|
CN104866815A (en) | 2015-08-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101334263B (en) | Circular target circular center positioning method | |
CN104915963B (en) | A kind of detection and localization method for PLCC elements | |
Grigorescu et al. | Contour and boundary detection improved by surround suppression of texture edges | |
CN107563377A (en) | It is a kind of to detect localization method using the certificate key area of edge and character area | |
CN106529448A (en) | Method for performing multi-visual-angle face detection by means of integral channel features | |
CN101809619A (en) | Physical property value estimating system, age estimating system, gender estimating system, age/gender estimating system, and physical property value estimating method | |
CN103500327B (en) | The model recognizing method of same brand vehicle based on spatial positional information | |
CN101625760A (en) | Method for correcting certificate image inclination | |
CN107066961B (en) | Fingerprint method for registering and device | |
CN106845542A (en) | Paper money number intelligent identification Method based on DSP | |
CN104866815B (en) | Fingerprint epipole accurate positioning method based on image spatial feature | |
CN102831403A (en) | Identification method based on fingerprint feature points | |
Mirza et al. | Design and implementation of indian paper currency authentication system based on feature extraction by edge based segmentation using Sobel operator | |
Shalaby et al. | A multilevel structural technique for fingerprint representation and matching | |
CN106778754A (en) | A kind of industrial ammeter digit recognition method of robust | |
CN104123554A (en) | SIFT image characteristic extraction method based on MMTD | |
Karna et al. | Normalized cross-correlation based fingerprint matching | |
Wang et al. | Automatic fundus images mosaic based on SIFT feature | |
CN105608689A (en) | Method and device for eliminating image feature mismatching for panoramic stitching | |
CN107895166B (en) | Method for realizing target robust recognition based on feature descriptor by geometric hash method | |
WO2017185507A1 (en) | Method and device for identifying multiple fingerprints | |
CN103886338A (en) | Multispectral image matching method and system thereof | |
CN106651756A (en) | Image registration method based on SIFT and authentication mechanism | |
CN103871048B (en) | Straight line primitive-based geometric hash method real-time positioning and matching method | |
CN103473546B (en) | A kind of Fingerprint diretion acquisition methods based on structure tensor |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20200826 Address after: 201104 floor 4, building 8, No. 168, Jixin Road, Minhang District, Shanghai Patentee after: Shanghai Zaide Information Technology Co.,Ltd. Address before: Hangzhou City, Zhejiang province 310018 Xiasha Higher Education Park No. 2 street Patentee before: HANGZHOU DIANZI University |
|
TR01 | Transfer of patent right |