CN100410964C - Acquisition and splicing method of three-face rolling fingerprint - Google Patents

Abstract

The present invention relates to a method for collecting and splicing three-surface rolling fingerprints. The method comprises the steps that a background image is collected; serial fingerprints are collected; the background of a fingerprint image is removed; a histogram of the obtained image is mapped into a histogram in which all the gray scales are uniformly distributed; the image is processed in a Gauss smoothing processing mode; the threshold value of the image is calculated, and the image is binarized; whether fingerprint input exists in the image is judged, if the fingerprint input does not exist in the image, the next image is collected, if the fingerprint input does not exist in the image, the next step is executed; the center and four boundaries of a fingerprint are calculated; the steps above are repeated, and the serial fingerprint image of the next image is collected; the center line of a most related block between the obtained serial fingerprint image and the previously obtained fingerprint image is an optimal splicing line; dislocation of the splicing line is eliminated; the two serial fingerprints are spliced; the steps above are repeated, the serial fingerprints are continuously collected and spliced until that the fingerprint input does not exist in one image. The three-surface rolling fingerprint collected by using the method is complete and has more characteristic points.

Description

The collection of three rolling fingerprints and joining method

Technical field

The present invention relates to a kind of fingerprint collecting and joining method, relate in particular to a kind of collection and joining method of three rolling fingerprints.

Background technology

Because the stability and the uniqueness of fingerprint itself make fingerprint play enormous function aspect identification, fingerprint collecting can be divided into plane fingerprint collecting and three rolling fingerprint collectings.

The plane fingerprint is applied in civilian occasion usually.Use under the situation of comparison in 1: 1 mostly, just be used for checking.During checking, the fingerprint that system is stored in the fingerprint of current stamp collection in the storehouse during with original registration is compared, the mass ratio that the fingerprint of storing in the storehouse is gathered when registering at that time better and be comparison in 1: 1, so in the technology that is used for verifying, the fingerprint feature point that needs compares less.Figure 13 is plane fingerprint acquisition instrument and the plane fingerprint that collects, its gatherer process is: finger is lain on the fingerprint acquisition instrument of plane, gather and generate a sheet of planar fingerprint image, resulting fingerprint image only is the center section of fingerprint generally, and the unique point that comprises is less relatively.

Three rolling fingerprints are applied in the alert occasion of using usually.Mostly use 1: under the situation of N comparison and N big especially, from several ten thousand to millions of, just be used for identification.The alert fingerprint that is commonly used to obtain with crime scene etc. with fingerprint compares.For example: do 1 with one piece of stamp fingerprint and fingerprint on site storehouse: the N comparison, it is relevant with which case to find out this suspect, and this process is called in technique of criminal investigation to be looked into; For another example: in the stamp fingerprint base, do 1 with one piece of fingerprint on site: the N comparison, which suspect finds out this case spot has occurred, and this process just is called in technique of criminal investigation to be looked into.Can see that with under the situation, because fingerprint on site is usually incomplete, of low quality, and be again 1: the N comparison, so require institute's print complete, fingerprint feature point is many alert.So the fingerprint of alert usefulness must be three rolling fingerprints.The mode of three rolling fingerprints of existing collection is to use the collection of printing ink stamp, be made into finger-print card and preservation then, but the fingerprint quality that uses this method to gather is relatively poor, fragile, preserve finger-print card and need set up huge finger-print card storehouse, and difficulty is also bigger when comparing.

Summary of the invention

The collection and the joining method that the purpose of this invention is to provide a kind of three rolling fingerprints, use the rolling fingerprint acquisition instrument, will point from left to right or rolling from right to left, in rolling process, gather a series of sequence fingerprints, and, generate three rolling fingerprints at last with these sequence fingerprint splicings.

In order to achieve the above object, the present invention adopts following technical scheme:

The collection of three rolling fingerprints of the present invention and joining method comprise the steps:

1) gathers Background, before finger beginning stamp, gather a frame Background earlier;

2) acquisition sequence fingerprint, finger rolls on acquisition window, gathers a width of cloth sequence fingerprint image in the rolling process;

3) fingerprint image goes background, and sequence fingerprint and the described Background that collects subtracted each other;

4) histogram with the resulting image of step 3) is mapped to the histogram that all uniform gray level distribute in the image;

5) image that step 4) is obtained is made Gauss's smoothing processing;

6) the image calculation threshold value that step 5) is obtained, and this image is carried out binaryzation according to threshold value;

7) determining step 6) whether have the fingerprint input in the image of gained, if do not have the fingerprint input to step 2) gather piece image down, have the fingerprint input then to arrive step 8);

8) center and four borders of the described fingerprint of calculating;

9) repeating step 2)-8), gather next width of cloth sequence fingerprint image, gathered behind the width of cloth to step 10);

10) just go out the piece of maximal correlation between sequence fingerprint image that step 9) obtains and the preceding sequence fingerprint image that once obtains, the center line of this piece is exactly best splicing line;

11) dislocation at elimination splicing line place is made smoothing processing to the sequence fingerprint border that will splice;

12) splicing two width of cloth sequence fingerprints, with the place ahead that is reversed of the direction of finger roll, last width of cloth sequence fingerprint is got the part of splicing line front, and a back width of cloth sequence fingerprint is got the part of splicing line back;

13) repeating step 9)-12), constantly gather and splice the sequence fingerprint, in judging a certain width of cloth image, do not had till the fingerprint input.

Owing to adopted technique scheme, three-sided fingerprint collection of the present invention and joining method can be gathered three rolling fingerprints rapidly accurately, and the fingerprint that collects is complete, and unique point is many, is easy to compare.

Description of drawings

Fig. 1 is the collection of three rolling fingerprints of the present invention and the process flow diagram of joining method;

Fig. 2 is a frame Background of gathering earlier before the finger beginning stamp in the method for the present invention;

Fig. 3 is a series of fingerprints that collect in the method for the present invention---a sequence fingerprint;

Fig. 4 is the design sketch that fingerprint image goes background in the method for the present invention;

Fig. 5 is a schematic diagram of in the method for the present invention image being made Gauss's smoothing processing;

Fig. 6 is the distribution plan of gradation of image after normalized in the method for the present invention;

Fig. 7 is the synoptic diagram that in the method for the present invention image is fitted;

Fig. 8 is a synoptic diagram of asking fingerprint image center and border in the method for the present invention;

Fig. 9 is a synoptic diagram of asking the maximal correlation piece of two width of cloth fingerprint images in the method for the present invention;

Figure 10 is a synoptic diagram of in the method for the present invention the splicing line border being made smoothing processing;

Figure 11 is the synoptic diagram that carries out the splicing of sequence fingerprint in the method for the present invention;

Figure 12 is three rolling fingerprints that in the method for the present invention a series of sequence fingerprints are spliced into;

Figure 13 is plane fingerprint acquisition instrument and the plane fingerprint that collects;

Figure 14 is three rolling fingerprint acquisition instruments and a series of sequence fingerprints that collect and spliced three rolling fingerprints.

Embodiment

Further specify technical scheme of the present invention below in conjunction with accompanying drawing.

Fig. 1 is the collection of three rolling fingerprints of the present invention and the process flow diagram of joining method, and as shown in Figure 1, the collection and the joining method of three rolling fingerprints of the present invention comprise:

1) gathers Background, before finger beginning stamp, gather a frame Background (S11) earlier;

2) acquisition sequence fingerprint, finger rolls on acquisition window, gathers a width of cloth sequence fingerprint image (S12) in the rolling process;

3) fingerprint image goes background, and sequence fingerprint and the described Background that collects subtracted each other (S13);

4) histogram with the resulting image of step 3) is mapped to the histogram (S14) that all uniform gray level distribute in the image;

5) image that step 4) is obtained is made Gauss's smoothing processing (S15);

6) the image calculation threshold value that step 5) is obtained, and this image is carried out binaryzation (S16) according to threshold value;

7) determining step 6) whether have the fingerprint input in the image of gained, if do not have the fingerprint input to step 2) gather piece image down, have the fingerprint input then to arrive step 8) (S17);

8) center and four borders (S18) of the described fingerprint of calculating;

9) repeating step 2)-8), gather next width of cloth sequence fingerprint image, gathered behind the width of cloth to step 10);

10) just go out the piece of maximal correlation between sequence fingerprint image that step 9) obtains and the preceding sequence fingerprint image that once obtains, the center line of this piece is exactly best splicing line (S19);

11) dislocation at elimination splicing line place is made smoothing processing (S110) to the sequence fingerprint border that will splice;

12) splicing two width of cloth sequence fingerprints, with the place ahead that is reversed of the direction of finger roll, last width of cloth sequence fingerprint is got the part of splicing line front, and a back width of cloth sequence fingerprint is got the part (S111) of splicing line back;

13) repeating step 9)-12), constantly gather and splice the sequence fingerprint, in judging a certain width of cloth image, do not had till the fingerprint input.

Step 1) is gathered Background, Fig. 2 is a frame Background of gathering earlier before the finger beginning stamp in the method for the present invention, the acquisition window of fingerprint acquisition instrument inevitably can be made dirty by dust, hand perspiration, greasy dirt etc., these factors all can cause when gathering fingerprint image and disturb and noise, influence the readability of fingerprint image, the escutcheon that stays when particularly gathering last piece of fingerprint is very big to the interference of fingerprint image.Before finger beginning stamp, gather a frame Background earlier, as shown in Figure 2, comprised original all traces on the acquisition window in this frame Background,, obtained current clear fingerprint image so that in follow-up step, from the collection fingerprint image, eliminate these markings.

Step 2) acquisition sequence fingerprint, Fig. 3 is a series of fingerprints that collect in the method for the present invention---a sequence fingerprint, finger is pressed down on the acquisition window of fingerprint acquisition instrument, finger rolls to the left side to the right side or from the right side from the left side, in the process of rolling, gather a series of fingerprint images successively---the sequence fingerprint, when handling, each width of cloth sequence fingerprint image that collects of handling.

The step 3) fingerprint image goes background, and Fig. 4 is the design sketch that fingerprint image goes background in the method for the present invention, with each width of cloth of the sequence fingerprint gathered all with background subtracting to dispel background.Concrete computing method are as follows: as shown in Figure 4, the wide N of the long M of fingerprint image, (i, gray-scale value j) they are that (i, j), (i, gray-scale value j) they are O to the sequence fingerprint to O more arbitrarily more arbitrarily for it ₁(i, j), Background more arbitrarily (i, gray-scale value j) be B (i, j), then: O ' (i, j)=O ₁(i, j)-B (i, j), O ' (i, j) ∈ [255,255]; With all gray values of pixel points O ' (i, j) do linear mapping: O ' (i, maximal value j) is max, minimum value is min, the gray-scale value after the linear mapping $0(i,j)=255*\frac{O^{\′}(i,j)-\min }{\max -\min },$ (i j) is the gray-scale value of dispeling after the background to O.What adopt in the present embodiment is the background that said method is removed fingerprint image, and said method is not unique method of removing background, and those of ordinary skill in the art can utilize additive method to remove background image.

The step 4) histogram equalization, when comparing two width of cloth images on an adhoc basis, general elder generation always carries out the regulation processing to the histogram of image, and making regulationization histogram afterwards is " standard histogram ".The most frequently used method of regulation processing is exactly a histogram equalization.It is by function b=f (a), and the histogram of original graph is mapped to the histogram that all uniform gray level distribute in the image, has increased the dynamic range of grey scale pixel value, thereby has reached the effect that strengthens the integral image contrast.For the suitable functions f (*) of expression input probability density and output probability density, f (*) is defined as follows:

$\mathrm{pb}\left(b\right)\mathrm{db}=\mathrm{pa}\left(a\right)\mathrm{da}\⇒\mathrm{df}=\mathrm{pa}\left(a\right)\mathrm{da}/\mathrm{pb}\left(b\right)$

F (*) is differentiable, and df/da 〉=0.Definition pb (b) is a constant, and P (a) is the intensity profile probability function, that is:

f(a)＝(2 ^B-1)*P(a)；

Step 5) is made Gauss's smoothing processing to image, and Gauss utilizes gaussian kernel to carry out picture smooth treatment, and it carries out neighborhood operation at image space by template and finishes, and allows image be suppressed in the high fdrequency component of fourier space, does not influence low frequency component.Because the gray-scale values such as edges of regions in the high fdrequency component correspondence image have the big very fast part that changes, wave filter can make image smoothing with these component eliminations, reaches the purpose that strengthens sharpness.Its key step is:

5.1) template is roamed in the drawings, and template center is overlapped with certain location of pixels among the figure;

5.2) respective pixel under coefficient on the template and the template is multiplied each other;

5.3) with all product additions

5.4) will compose the pixel of giving corresponding templates center among the figure with (the output response of template).

Fig. 5 is a schematic diagram of in the method for the present invention image being made Gauss's smoothing processing, and wherein (a) is the part of piece image, wherein the gray-scale value of some pixels that are designated as.Now establish a 3*3 template, shown in (b), the coefficients that is designated as in the template.As with k ₀Gray-scale value is s among position and the figure ₀Pixel overlap, the output response R of template is: R=k ₀s ₀+ k ₁s ₁+ ... k ₈s ₈

R is composed to enhancing figure, as at (x, gray-scale value y).See (c)

To all operations like this of each pixel of former figure, the new gray-scale value of all positions of figure that are enhanced.Just can obtain different high passes or low pass effect for the different value of k tax in the template.

Gauss uses Gaussian function $h(x,y)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}e-((x^2+y^2)/2{\mathrm{\σ}}^2)$ H ' after the normalization (x y) does the neighborhood operation of template, and normalized is as follows: $h^{,}(x,y)=\frac{k_i}{\underset{i=1}{\overset{9}{\mathrm{\Σ}}}{k}_i}$

What adopt in the present embodiment is that said method comes image is carried out Gauss's smoothing processing, and said method is not unique method of carrying out Gauss's smoothing processing, and those of ordinary skill in the art can utilize additive method to come image is carried out Gauss's smoothing processing.

Step 6) is to the image calculation threshold value, and according to threshold value this image carried out binaryzation, does binaryzation and asks threshold value in the past earlier.The fingerprint image of a similar Fig. 7, any peak in bimodal all is normal distribution,

Threshold value thr should be adaptive.Now hypothesis think gray-scale value less than 10 do not have a fingerprint image input.Rule of thumb estimate б=20, then with test

H：h(i)，i∈[0，255]；

AH：ah(0)＝h(0)；

ah(i)＝ah(i-1)+h(i)；

Fig. 7 is the synoptic diagram that in the method for the present invention image is fitted, and obtains a pair of (x by match ₁, б ₁), (x ₂, б ₂), make the error of fitting minimum.Obtain threshold value thr=(x so ₁+ x ₂)/2;

Image is carried out binaryzation, and for any gray-scale value of any in the image, order all equals 255 more than or equal to the gray scale of threshold value, all is 0 less than the gray scale of threshold value, that is:

I (i, j)=255 work as I ' (i, j) 〉=during thr

I (i, j)=0 work as I ' (i, j)＜during thr;

What adopt in the present embodiment is that said method and parameter come image is carried out binary conversion treatment, said method and selected parameter are not unique, and those of ordinary skill in the art can utilize additive method or select different parameters to come image is carried out binary conversion treatment.

Step 7) judges that fingerprint has or not, and judges fingerprint according to the threshold value Thr that asks in the last step and has or not, and gray scale has been judged as fingerprint input during greater than threshold value.

Step 8) is asked the center and four borders of sequence fingerprint, the fingerprint that collects is judged four borders and the center of each two field picture in the sequence fingerprint after binaryzation, Fig. 8 is a synoptic diagram of asking fingerprint image center and border in the method for the present invention, 81 is sweep trace among the figure, 82 is the fingerprint image border, 83 is the direction of scanning, and 84 is the fingerprint frontier point.

Step 10) is asked best splicing line and splicing width, and Fig. 9 is a synoptic diagram of asking the maximal correlation piece of two width of cloth fingerprint images in the method for the present invention, to the two continuous frames fingerprint image that collects, asks the piece of maximal correlation earlier, and the size of establishing fingerprint image is M*N,

MaxCorr＝Max(Corri)i＝1，…n

Index＝MaxCorr＝Corrindex

Ask relevant:

I=0 ~ M-1, j=0 ~ N-1;

Obtain the piece of maximal correlation, best splicing line is exactly the center line of piece, and 91 is back one sequence fingerprint among the figure, and 92 is the maximal correlation piece, and 93 for the maximal correlation center line is a splicing line, and 94 is last sequence fingerprint.

What adopt in the present embodiment is the piece that said method is asked maximal correlation between the two continuous frames fingerprint image, and said method is not unique, and those of ordinary skill in the art can utilize additive method to ask the piece of maximal correlation between the two continuous frames fingerprint image.

Step 11) is eliminated the dislocation at splicing line place, smoothing processing is made on the sequence fingerprint border that will splice, Figure 10 is a synoptic diagram of in the method for the present invention the splicing line border being made smoothing processing, the splicing line place of sequence fingerprint is a needle prick shape as shown in figure 10, and earlier smoothing processing is made on sequence fingerprint border before the splicing: (x sets up an office _i, y _i) be that certain fingerprint is arbitrarily borderline, then, $x_i^{\′}=\underset{k=i-s}{\overset{i+s}{\mathrm{\Σ}}}{x}_k/(2s+1)$ $y_i^{\′}=\underset{k=i-s}{\overset{i+s}{\mathrm{\Σ}}}{y}_k/(2s+1)$ S is a smoothing factor, (x ' _i, y ' _i) be the border after level and smooth.

What adopt in the present embodiment is that said method is eliminated the dislocation at splicing line place and smoothing processing is made on the sequence fingerprint border that will splice, said method is not unique, and those of ordinary skill in the art can utilize additive method to eliminate the dislocation at splicing line place and smoothing processing is made on the sequence fingerprint border that will splice.

Step 12) is spliced two width of cloth sequence fingerprints, the place ahead that is reversed with the direction of finger roll, sequence fingerprint image front one frame is got the part of center line (the being splicing line) front of maximal correlation piece, a next frame is got the part of center line (the being splicing line) back of maximal correlation piece, and Figure 11 is the synoptic diagram that carries out the splicing of sequence fingerprint in the method for the present invention.

Do continuous splicing up to judging that finger leaves, in whole gatherer process, first piece of fingerprint got first piece and the preceding whole fingerprint images of second piece of sequence fingerprint maximal correlation piece center line (being splicing line), except that last piece fingerprint, the maximal correlation piece part fingerprint image that other sequence fingerprints are got behind the maximal correlation piece center line (being splicing line) splices on last piece of fingerprint, repeat above-mentioned each step, just with whole fingerprint image splicings three good rolling fingerprints of generation splicing on last piece of sequence fingerprint of last piece sequence fingerprint maximal correlation piece center line back, Figure 12 is three rolling fingerprints that in the method for the present invention a series of sequence fingerprints are spliced into when judging no fingerprint.

Figure 14 is three rolling fingerprint acquisition instruments and a series of sequence fingerprints that collect and spliced three rolling fingerprints, and as seen from the figure, spliced three rolling fingerprints are compared to the plane fingerprint and seem more complete, and unique point is also more.

Claims (7)

1. the collection and the joining method of three rolling fingerprints comprise the steps:

1) gathers Background, before finger beginning stamp, gather a frame Background earlier;

2) acquisition sequence fingerprint, finger rolls on acquisition window, gathers a width of cloth sequence fingerprint image in the rolling process;

3) fingerprint image goes background, and sequence fingerprint and the described Background that collects subtracted each other;

4) histogram with the resulting image of step 3) is mapped to the histogram that all uniform gray level distribute in the image;

5) image that step 4) is obtained is made Gauss's smoothing processing;

6) the image calculation threshold value that step 5) is obtained, and this image is carried out binaryzation according to threshold value;

7) determining step 6) whether have the fingerprint input in the image of gained, if do not have the fingerprint input to step 2) gather piece image down, have the fingerprint input then to arrive step 8);

8) center and four borders of the described fingerprint of calculating;

9) repeating step 2)-8), gather next width of cloth sequence fingerprint image, gathered behind the width of cloth to step 10);

10) obtain the piece of maximal correlation between sequence fingerprint image that step 9) obtains and the preceding sequence fingerprint image that once obtains, the center line of this piece is exactly best splicing line;

11) dislocation at elimination splicing line place is made smoothing processing to the sequence fingerprint border that will splice;

12) splicing two width of cloth sequence fingerprints, with the place ahead that is reversed of the direction of finger roll, last width of cloth sequence fingerprint is got the part of splicing line front, and a back width of cloth sequence fingerprint is got the part of splicing line back;

13) repeating step 9)-12), constantly gather and splice the sequence fingerprint, in judging a certain width of cloth image, do not had till the fingerprint input.

2. the collection of three rolling fingerprints as claimed in claim 1 and joining method is characterized in that, described 3) fingerprint image goes the method for background as follows:

The wide N of the long M of original series fingerprint image that gathers, (i, gray-scale value j) are that (i, j), (i, gray-scale value j) are O to the actual sequence fingerprint to O more arbitrarily more arbitrarily for it ₁(i, j), Background more arbitrarily (i, gray-scale value j) be B (i, j), then: O ' (i, j)=O ₁(i, j)-B (i, j), O ' (i, j) ∈ [255,255], with all gray values of pixel points O ' (i, j) do linear mapping: O ' (i, maximal value j) is max, minimum value is min, the gray-scale value after the linear mapping $0(i,j)=255*\frac{O^{\′}(i,j)-\min }{\max -\min },$ (i j) is the gray-scale value of dispeling the sequence fingerprint after the background to O.

3. the collection of three rolling fingerprints as claimed in claim 1 and joining method is characterized in that, described 4) histogram is mapped to the histogram that all uniform gray level distribute in the image and is meant by function b=f (a) f (a)=(2 ^B-1) * P (a) thus with histogram be mapped in the image histogram that all uniform gray level distribute increase grey scale pixel value dynamic range, strengthen the integral image contrast.

4. the collection of three rolling fingerprints as claimed in claim 1 and joining method is characterized in that, described 5) image being made Gauss's smoothing processing may further comprise the steps:

5.1) template is roamed in the drawings, and template center is overlapped with certain location of pixels among the figure;

5.2) respective pixel under coefficient on the template and the template is multiplied each other;

5.3) with all product additions;

5.4) will compose the pixel of giving corresponding templates center among the figure with (the output response of template).

5., and according to threshold value this image is carried out binaryzation and comprises the collection of three rolling fingerprints as claimed in claim 1 and joining method is characterized in that, described 6) to the image calculation threshold value:

The gray-scale value of the fingerprint image after Gauss's smoothing processing is normal distribution, utilizes formula

ah(0)＝h(0)；

ah(i)＝ah(i-1)+h(i)；

H (i) is a Gaussian function, i ∈ [0,255];

Carry out match, obtain a pair of (x ₁, δ ₁) (x ₂, δ ₂), make to fit the error minimum,

Calculated threshold thr=(x ₁+ x ₂)/2;

According to described threshold value thr image is carried out binaryzation, to gray scale greater than threshold value thr gray scale is made as 255, gray scale is made as 0 less than threshold value thr with gray scale.

6. the collection of three rolling fingerprints as claimed in claim 1 and joining method is characterized in that, described 10) piece of maximal correlation is meant between sequence fingerprint image that obtains and the preceding sequence fingerprint image that once obtains: the size of establishing fingerprint image is M*N,

MaxCorr＝Max(Corr ₁)i＝1，…n

Index＝MaxCorr＝Corr _index

Ask relevant:

I=0 ~ M-1, j=0 ~ N-1;

Obtain the piece of maximal correlation, best splicing line is exactly the center line of piece.

7. the collection of three rolling fingerprints as claimed in claim 1 and joining method is characterized in that, described 11) eliminate the dislocation at splicing line place, the sequence fingerprint border that splice is meant as smoothing processing: (x sets up an office _i, y _i) be that certain fingerprint is arbitrarily borderline, then $x_i^{\′}=\underset{k=i-s}{\overset{i+s}{\mathrm{\Σ}}}{x}_k/(2s+1)$ $y_i^{\′}=\underset{k=i-s}{\overset{i+s}{\mathrm{\Σ}}}{y}_k/(2s+1)$ S is a smoothing factor, (x ' _i, y ' _i) be the border after level and smooth.

Images