CN105632015A - Bank bill fingerprint feature anti-counterfeiting identification method based on cloud platform - Google Patents

Abstract

The invention discloses a bank bill fingerprint feature anti-counterfeiting identification method based on a cloud platform. The method comprises the following steps that S1, multispectral images of a big bill are acquired; S2, the acquired multispectral images are preprocessed so that standard single bill images are obtained, and the fingerprint features of all the single bill images are stored in the cloud platform; S3, image information of a bill to be detected is acquired, and the acquired image information is respectively detected through double color fiber detection, fluorescent pattern detection and infrared image blanking feature detection with combination of the fingerprint features of the single bill images stored in the cloud platform; and S4, when the bill to be detected is judged to be true through three types of detection, the bill to be detected is judged to be true, or the bill to be detected is judged to be false. According to the method, the defects of anti-counterfeiting identification through the single bill identification technology can be eliminated so that accuracy of anti-counterfeiting identification can be effectively enhanced.

Description

A kind of bank money fingerprint characteristic anti false authentication method based on cloud platform

Technical field

The present invention relates to a kind of financial document anti false authentication method, particularly relate to a kind of bank money fingerprint characteristic anti false authentication method based on cloud platform, belong to anti-counterfeit and counterfeit distinguishing technical field.

Background technology

Along with the fast development of national economy, the application of financial document is more and more extensive. But at present financial document in management, use and also there are some problems in false distinguishing, the lawless person in society has directed pointing crime target bank, and financial document is swindled case and happened occasionally, and causes great financial loss to country. Existing financial document false distinguishing method mainly relies on the problem that artificial qualitative analysis is main, existing to be manually differentiate that intensity is big, for a long time consuming time, and easily owing to tired or carelessness cause inspection by mistake.

Fiber is one of important anti-false sign of bill, and in different bill, the stochastic distribution of position of fibers is different. Therefore, in existing bill anti-counterfeit discrimination method, usually adopt extraction fiber characteristics to carry out bill anti-counterfeit discriminating. Under UV-irradiation, fiber has fluorescent effect, is convenient to extract anti-counterfeiting characteristic. But, in this false proof discriminating process, it is committed step and a difficult problem that the fiber characteristics of bill surface extracts. Owing to the background of bill surface is very complicated, comprise word, frame, fluorized marking and shoulder etc., intensity profile wide range, and fiber target is smaller, although gray scale is roughly distributed in high clear zone, but do not have obvious boundary with background intensity profile, distinguish more difficult. And in putting into practice, requiring the machine extraction fiber characteristics similar for same bill structure, its result wants consistent, and this just requires that the adaptivity of feature extracting method and stability all to be got well.

In order to solve the problem, it is that the Chinese invention patent of ZL201110362933.1 proposes a kind of bill anti-counterfeit discrimination method based on fiber individualized feature in the patent No., comprise the steps: that the bill images got by pick up camera processes, obtain the bill images of stdn; Adopt maximum filter to carry out image filtering, the multiclass object in bill images is changed into shoulder background and fiber target two class object; Adopt the two-dimensional entropy segmentation bill images optimized, detection fibers target; Extracting the anti-counterfeiting characteristic of fiber target, anti-counterfeiting characteristic is one or more in center-of-mass coordinate, area, curvature and moment characteristics; Carry out characteristic matching based on anti-counterfeiting characteristic, differentiate the true and false of bill. The method has merged maximum value filtering and has improved the fiber small target deteection technology of two-dimensional entropy, has good adaptivity and stability.

Although aforesaid method can improve the accuracy of false proof discriminating to a certain extent. But, along with the development and progress of science and technology, the level that illegal molecule makes fictitious bill also changes along with the appearance of new edition bill, and fraud level is also more and more higher. Single bill authentication technique is difficult to the accuracy ensureing bill carries out false proof qualification.

Summary of the invention

For the deficiencies in the prior art, technical problem to be solved by this invention is to provide a kind of bank money fingerprint characteristic anti false authentication method based on cloud platform.

For achieving the above object, the present invention adopts following technical scheme:

Based on a bank money fingerprint characteristic anti false authentication method for cloud platform, comprise the steps:

S1, gathers the multispectral image of a big bill;

S2, carries out pre-treatment to described multispectral image, obtains the sola bill image of standard, and the fingerprint characteristic of all sola bill images is stored into cloud platform;

S3, gathers the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, described graphic information is detected respectively by double-colored fiber check and measure, phosphor pattern detection, infrared image blanking feature detection;

S4, when three kinds of detections all judge that bill to be detected is as, time true, judging that bill to be detected is as true, otherwise, judge that bill to be detected is as vacation.

Wherein more preferably, in step s 2, described multispectral image is carried out pre-treatment, obtain the sola bill image of standard, comprise the steps:

S21, according to distortion factor and distortion model, carries out distortion correction to the multispectral image of the big bill gathered;

S22, to the image carrying out obtaining after distortion corrects, carries out the sola bill image that scaling cutting obtains standard.

Wherein more preferably, in step S22, to the image carrying out obtaining after distortion corrects, carry out the sola bill image that scaling cutting obtains standard, comprise the steps:

S221, for the white light reflection image WIMG of a big bill, is transformed into hsv color space by RGB color;

S222, adopt fixed threshold dividing method, using H, S component in white light reflection image WIMG respectively the pixel in the first fixed interval as object point, obtain binary segmentation image;

S223, adopts the rectangular window identical with bill water solvus size to be scanned by image, the object point number in statistics rectangular window, if object point data exceed the 1/3 of rectangular window size, water solvus position then being detected, record rectangular window central position is (x0, y0);

S224, with coordinate point (x0-w1, y0-h1) it is bill starting point, the cutting of sola bill image is carried out according to bill size, obtain the sola bill image of standard, wherein, w1 is the distance on the left of bill water solvus regional center distance bill, and h1 is the distance on the upside of bill water solvus regional center distance bill.

Wherein more preferably, in step s3, gather the graphic information of bill to be detected, by double-colored fiber check and measure, the graphic information gathered is detected, comprise the steps:

S301, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and converts thereof into the picture of HSV form;

S302, according to the feature of blue fiber red in bill, using H, S component in the picture of HSV form respectively the pixel in the 2nd fixed interval as object point, carry out H, S component coarse segmentation, V component adopts OTSU Threshold Segmentation Algorithm carry out image subdivision and cuts;

S303, carries out fiber target extraction to the figure after segmentation;

S304, obtain the fiber that same bar fiber presents respectively in the ultraviolet reflectance image FIMG in front and the ultraviolet reflectance image BIMG at the back side, fiber in the ultraviolet reflectance image FIMG in front and the ultraviolet reflectance image BIMG at the back side is mated, if two bar fiber mates mutually, bill to be detected is by double-colored fiber check and measure; Otherwise, bill to be detected is false tickets.

Wherein more preferably, when carrying out the true and false of double-colored fiber check and measure bill to be detected, the true and false of bill to be detected is adjudicated further by double-colored characteristic.

Wherein more preferably, the described true and false adjudicating bill to be detected by double-colored characteristic comprises the steps:

For same bar fiber, extract the minimum value of its H component in the ultraviolet reflectance image FIMG in front and the ultraviolet reflectance image BIMG at the back side, mean value and maximum value respectively, it is designated as respectively

The double-colored characteristic judgement function built is:

Judge the value of F; If F is not zero, then described fiber meets double-colored characteristic, and bill to be detected is true; Otherwise, described fiber does not meet double-colored characteristic, and bill to be detected is false.

Wherein more preferably, in step s3, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, by phosphor pattern detection, the graphic information gathered is detected, comprise the steps:

S311, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and converts thereof into the picture of HSV form;

S312, to the V component of the ultraviolet reflectance image FIMG in front, carries out little wave conversion, obtains the low-frequency information FIMG0 after conversion;

S313, to the V component of described low-frequency information FIMG0, adopts OTSU Threshold Segmentation Algorithm to carry out image subdivision and cuts;

S314, extracts the fingerprint characteristic of sola bill image of standard from cloud platform, and the phosphor pattern feature of image after segmentation being cut is mated with the phosphor pattern feature of the sola bill image of standard, true and false by the value decides bill of relation conefficient.

Wherein more preferably, in step S314, the calculation formula of described relation conefficient is:

Wherein, the low-frequency information of the ultraviolet reflectance image FIMG that FIMG0 (x, y) is front on V component; The two-value template image of the sola bill image that MIMG (x, y) is standard; When relation conefficient is greater than phosphor pattern detection threshold, judge that phosphor pattern meets false proof requirement; Bill is true; Otherwise, judging that phosphor pattern does not meet false proof requirement, bill is false.

Wherein more preferably, in step s3, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, by infrared image blanking feature detection, the graphic information gathered is detected, comprise the steps:

S321, gathers the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected respectively, and is converted to gray-scale map picture;

S322, to the gray-scale map picture of the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected, carries out little wave conversion respectively, obtains the low-frequency information RIMG0 after conversion and WIMG0;

S323, builds corresponding threshold binary image respectively, carries out threshold binary image segmentation low-frequency information RIMG0 and WIMG0, obtains corresponding binary map as RIMG1 and WIMG1;

S324, merges binary map as RIMG1 and WIMG1, obtains binary map as RWIMG1:

S325, extract the two-value template image of the infrared image of the sola bill image of standard from cloud platform, adopt correlation matching algorithm to be mated with described two-value template image by described binary map picture, if the match is successful, then the infrared blanking characteristic of band detection bill meets false proof requirement, and bill is true; Otherwise, judging that the infrared blanking characteristic of band detection bill does not meet false proof requirement, bill is false.

Wherein more preferably, in step S323, low-frequency information RIMG0 is built threshold binary image, comprises the steps:

For any pixel point (x, y), calculate the gray average M1 in its neighborhood window;

Judge whether inequality R IMG0 (x, y) > M1 sets up, if set up, threshold binary image TIMG (x, the y)=M1 at point (x, y) place;

Otherwise, calculate optimal segmenting threshold t, threshold binary image TIMG (x, the y)=t at point (x, y) place.

Bank money fingerprint characteristic anti false authentication method based on cloud platform provided by the present invention, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, respectively the graphic information gathered is detected by double-colored fiber check and measure, phosphor pattern detection, infrared image blanking feature detection; When three kinds of detections all judge that bill to be detected is as, time true, judging that bill to be detected is as true, when three kinds of detections have one to judge bill to be detected as vacation, judge that bill to be detected is as vacation. Eliminate the drawback that single bill authentication technique carries out false proof qualification, effectively improve the accuracy of false proof qualification.

Accompanying drawing explanation

Fig. 1 is the schema of the bank money fingerprint characteristic anti false authentication method based on cloud platform provided by the present invention;

Fig. 2 is the circuit connection diagram that the multi-optical spectrum image collecting system of bill is opened greatly by bank;

Fig. 3 is in anti false authentication method provided by the present invention, and employing high-definition network camera gathers the structural representation that the multispectral image of draft is opened greatly by bank;

Fig. 4 is in anti false authentication method provided by the present invention, the position distribution schematic diagram of separate unit pick up camera and light source, draft;

Fig. 5 is in anti false authentication method provided by the present invention, pixel grey scale distribution histogram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the technology contents of the present invention is carried out detailed specific description.

As shown in Figure 1, first the bank money fingerprint characteristic anti false authentication method based on cloud platform provided by the present invention, comprised the steps:, opened greatly the multispectral image of bill by camera acquisition; Secondly, the multispectral image of the big bill gathered is carried out pre-treatment, obtains the sola bill image of standard, and the fingerprint characteristic of all sola bill images is stored into cloud platform; Then, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, respectively the graphic information of the bill to be detected gathered is detected by double-colored fiber check and measure, phosphor pattern detection, infrared image blanking feature detection; When three kinds of detections all judge that bill to be detected is as, time true, judging that bill to be detected is as true, when three kinds of detections have one to judge bill to be detected as vacation, judge that bill to be detected is as vacation. Wherein, in embodiment provided by the present invention, the fingerprint characteristic of bank money comprises the double-colored fiber characteristics of bank money, phosphor pattern feature and infrared image blanking feature etc. This process is done detailed specific description below.

S1, opens greatly the multispectral image of bill by camera acquisition.

In embodiment provided by the present invention, a big bill is made up of 20 Standard Bank bills, is opened greatly the multispectral image of bill by camera acquisition, is the multispectral image of the big draft adopting 4 high-definition network cameras to gather bank. Or adopt CMOS line-scan digital camera module collection to open greatly the multispectral image of bill. In embodiment provided by the present invention, the big multispectral image opening draft gathering bank for 4 high-definition network cameras is described. Wherein, the multi-optical spectrum image collecting system that the multispectral image gathering a big bill of bank with 4 high-definition network cameras adopts bank to open greatly bill realizes. The circuit that the multi-optical spectrum image collecting system of bill is opened greatly by bank connects as shown in Figure 2.

Hardware device mainly comprises ARM switchboard, light source module, high-definition network camera (pick up camera) three parts. Wherein, ARM switchboard is mainly used in Energy control, powers on to different light sources (infrared light, UV-light and white light) and 4 high-definition network cameras according to the order of front-end computer. ARM version have employed LPC2131 chip, and the LPC2131 microcontroller adopting LPC2131 chip is the ARM7TDMI-SCPU supporting real-time simulation and Embedded Trace based on.

When selecting high-definition network camera, in order to gather the bank money image of high-resolution, adopt the high-resolution pick up camera of ten million Pixel-level. In addition, owing to needing the diffuse reflectance infrared spectroscopy gathering image, it is necessary to the 700nm low pass colour filter of removal camera front end.

In order to gather multispectral bank money image, light source module adopts three kinds of LED/light source, comprises infrared, ultraviolet source and white light, and three kinds of light sources are integrated on a light source board, have independent 12V power supply respectively, control by ARM switchboard, can power on individually or simultaneously.

To adopting 4 high-definition network cameras to gather, multispectral image that bank opens greatly bill is described below. Pickup area is illustrated as shown in Figure 3. Wherein 2 pick up cameras in top respectively gather 4 width bill images, and 2, lower section pick up camera respectively gathers 6 width bill images.

Light source comprises infrared light, UV-light and white light, it is divided into upper and lower two groups, the position distribution of separate unit pick up camera and light source, bill is as shown in Figure 4, one group of light source board provides light source above, for gathering bill reflection spectrum images, one group of light source board provides light source below, for gathering bill transmitted spectrum image. By the conversion of not sharing the same light of infrared light, UV-light and white light, it is possible to gather the graphic information under the light source irradiation needed.

S2, carries out pre-treatment to the multispectral image of the big bill gathered, obtains the sola bill image of standard, and the fingerprint characteristic of all sola bill images is stored into cloud platform.

After by camera acquisition to the big multispectral image opening bill, owing to camera lens generally exists distortion, and when carrying out the detection of bill to be detected, need the fingerprint characteristic with sola bill to check, so needing the multispectral image to the big bill gathered to carry out pre-treatment, obtain the sola bill image of standard, and extract the fingerprint characteristic of sola bill image, the fingerprint characteristic of all sola bill images is stored into cloud platform, in order to the offer of data during follow-up veritification. Wherein, the multispectral image of the big bill gathered is carried out pre-treatment, obtains the sola bill image of standard, specifically comprise the steps:

S21, according to distortion factor and distortion model, carries out distortion correction to the multispectral image of the big bill gathered.

Generally there is distortion in camera lens, it is thus desirable to the multispectral image of the big bill gathered is carried out distortion correction. A big bill gathers the distortion of camera lens based on radial distortion, and distortion degree is little. Therefore taking a rank radial distortion as distortion model, represent and be:

$\left\{\begin{array}{c}{x}_u=x_d(k_0+k_1{r_d}^2)\\ y_u=y_d(k_0+k_1{r_d}^2)\end{array}\right.$

Wherein, x_u,y_uFor ideal image coordinate; x_d, y_dFor fault image coordinate, r_dRepresent point (x_d,y_d) with the distance of true origin, namelyk₀��k₁For distortion factor.

Upper formula can be rewritten as:

$\left\{\begin{array}{c}\frac{x_u}{x_d}=k_0+k_1{r_d}^2\\ \frac{y_u}{y_d}=k_0+k_1{r_d}^2\end{array}\right.$

Orderr_d ²=x, then distortion model can represent and is:

Y=k₀+k₁x

Upper formula has two unknown distortion parameters. In order to ask for this two parameters, adopt checker to be demarcated by pick up camera, obtain n to reference mark { (x_i,y_i); I=1,2 ..., n}, In conjunction with distortion model, adopt cumulative sum method, build group of functions:

$\left\{\begin{array}{c}{\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}{y}_i=k_0{\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}+k_1{\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}{x}_i\\ {\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}{y}_i=k_0{\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}+k_1{\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}{x}_i\end{array}\right.$

Wherein, subscript (1) and (2) represent 1 rank and 2 rank Cumulate Sum respectively, and k rank Cumulate Sum is defined as

${\underset{i=1}{\overset{n}{\Σ}}}^{\left(k\right)}{x}_i=\underset{i=1}{\overset{n}{\Σ}}{\underset{j=1}{\overset{i}{\Σ}}}^{(k-1)}{x}_j$

AndRepresenting the basic Cumulate Sum in k rank, calculation formula is

${\underset{i=1}{\overset{n}{\Σ}}}^{\left(k\right)}=\frac{1}{k!}n(n+1)...(n+k-1)$

If $\mathrm{\Δ}=\left|\begin{array}{cc}{\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}& {\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}{x}_i\\ {\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}& {\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}{x}_i\end{array}\right|\≠0,$ Can obtain distortion factor is:

$\left[\begin{array}{c}{k}_0\\ k_1\end{array}\right]={\left[\begin{array}{cc}{\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}& {\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}{x}_i\\ {\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}& {\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}{x}_i\end{array}\right]}^{-1}\left[\begin{array}{c}{\underset{i=1}{\overset{n}{\Σ}}}^{\left(1\right)}{y}_i\\ {\underset{i=1}{\overset{n}{\Σ}}}^{\left(2\right)}{y}_i\end{array}\right]$

According to distortion factor and distortion model, carry out distortion correction. For fault image coordinate (x_u,y_u), corresponding calibration coordinate (x_d,y_d) it is

$\left\{\begin{array}{c}{x}_d=x_u\frac{r_d}{r_u}\\ y_d=y_u\frac{r_d}{r_u}\end{array}\right.$

Wherein, rd is point (x_u,y_u) with the distance of true origin, namelyThe i.e. big distance of multispectral image coordinate to true origin opening bill, r_dRepresent point (x_d,y_d) with the distance of true origin, namely calibration coordinate is to the distance of true origin;

r_dMeet following condition:

$r_u=(k_0+k_1{r_d}^2)\sqrt{{x_d}^2+{y_d}^2}=r_d(k_0+k_1{r_d}^2)$

For the coordinate (x after correction_d,y_d), calculate (x on the rear image of correction by bilinear interpolation method_u,y_u) gray-scale value put.

S22, to the image carrying out obtaining after distortion corrects, carries out the sola bill image that scaling cutting obtains standard.

Consider that bill water solvus position is relatively fixing, and it is easy to detection. Therefore, the present invention calibrates according to bill water solvus position, carries out image cropping according to the priori of bill size. To the image carrying out obtaining after distortion corrects, carry out the sola bill image that scaling cutting obtains standard, specifically comprise the steps:

S221, for the white light reflection image WIMG of a big bill, is transformed into hsv color space by RGB color, and conversion formula is as follows:

$\left[\begin{array}{c}V\\ V_1\\ V_2\end{array}\right]=\left[\begin{array}{ccc}\frac{1}{3}& \frac{1}{3}& \frac{1}{3}\\ \frac{-1}{\sqrt{6}}& \frac{-1}{\sqrt{6}}& \frac{2}{\sqrt{6}}\\ \frac{1}{\sqrt{6}}& \frac{-1}{\sqrt{6}}& 0\end{array}\right]\left[\begin{array}{c}R\\ G\\ B\end{array}\right];$

$H=arctan\left\{\frac{V_2}{V_1}\right\};$

$S={({V_1}^2+{V_2}^2)}^{\frac{1}{2}};$

Wherein, RGB is a kind of color standard of industry member, it is by the change of red (R), green (G), blue (B) three Color Channels and their superpositions each other are obtained color of all kinds, namely RGB is the color representing red, green, blue three passages, this standard almost include human eyesight can all colours of perception, be use one of the widest color system at present. HSV (Hue, Saturation, Value) being a kind of color space, also claim hexagonal pyramid model (HexconeModel), in this model, the parameter of color is respectively: tone (H), saturation ratio (S), brightness (V). V₁And V₂For the middle variable of brightness.

S222, adopt fixed threshold dividing method, using H, S component in white light reflection image WIMG respectively the pixel in the first fixed interval as object point, obtain binary segmentation image; Wherein, in white light reflection image WIMG, the first fixed interval of H, S component is respectively [270,360], [0.0,0.2].

S223, adopts the rectangular window being of a size of 950 �� 160 to be scanned by image, and wherein, in the bill images that acquisition system gathers, the width in water solvus region is about 950, is highly about 160. Object point number in statistics rectangular window, if object point data exceed the 1/3 of rectangular window size, then thinks and water solvus position detected, now recording rectangular window central position is (x0, y0);

S224, carries out the cutting of sola bill image according to the priori of bill size, and bill starting point coordinate is (x0-w1, y0-h1), and cutting image is of a size of 2048 �� 1536. Wherein, w1 is the distance on the left of bill water solvus regional center distance bill, and h1 is the distance on the upside of bill water solvus regional center distance bill, in embodiment provided by the present invention, and w1=760, H1=590. In the bill images that acquisition system gathers, the width in sola bill region is about 2048, is highly about 1536.

According to above-mentioned steps, carrying out the image obtained after distortion correction is syncopated as all sola bill images from a big bill, and all sola bill images and its fingerprint characteristic comprised are stored into cloud platform, effectively ensure that the accuracy of the sola bill image of standard and comprehensive, improve the comprehensive of anti-counterfeiting detection Data Source, and then improve the accuracy of detection, it is stored into cloud platform and effectively saves local storage space.

S3, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, respectively the graphic information of the band detection bill gathered is detected by double-colored fiber check and measure, phosphor pattern detection, infrared image blanking feature detection.

When bill to be detected is carried out authenticity verification, the detection of double-colored fiber check and measure, phosphor pattern, infrared image blanking feature detection is adopted the graphic information gathered to be detected respectively, only when three kinds of detections all judge that bill to be detected is as time true, judge that bill to be detected is as true, when three kinds of detections have a kind of detection to judge bill to be detected as vacation, judge that bill to be detected is as vacation. The order of three kinds of detections can adjust as required, is described respectively below.

Gather the graphic information of bill to be detected, by double-colored fiber check and measure, the graphic information gathered is detected, specifically comprise the steps:

S301, control light source, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and converts thereof into the picture of HSV form.

Control light source, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and the image of collection is RGB24 form, owing to RGB component is subject to light-source brightness impact, therefore is converted into hsv color space.

S302, according to the feature of blue fiber red in bill, using H, S component in the picture of HSV form respectively the pixel in the 2nd fixed interval as object point, carry out H, S component coarse segmentation, V component adopts OTSU Threshold Segmentation Algorithm carry out image subdivision and cuts.

Fiber in 2010 editions bills has red blue two kinds of colors, and statistics finds, its H component is respectively in interval [270,360] and [160,240], and S component is all in interval [0.3,0.6]. For this reason, adopt fixed threshold method here, using the interval [0.3,0.6] of the interval [270,360] of H component and [160,240] and S component as the 2nd fixed interval, the HS component of image FIMG and BIMG is carried out coarse segmentation.

Specifically, any pixel (i, j) of zone of fiber meets the following conditions:

$\left\{\begin{array}{c}0.3\≤S(i,j)\≤0.6\\ 160\≤H(i,j)\≤240\end{array}\right.$ Or 270��H (i, j)��360

After Iamge Segmentation, target pixel value is designated as 1, and background pixel value is designated as 0, lower same.

In HS component coarse segmentation process, it is loose that threshold value arranges comparison, although like this can not missing inspection fiber pixel, but part non-fiber pixel can be examined by mistake.

Statistics finds, on V component, the general brightness of zone of fiber is bigger. Although in entire image, the pixel intensity of part background area is also relatively big, but in the neighborhood near fiber, the brightness of fiber pixel and near it luminance difference of background pixel bigger. For this reason, here in each zone of fiber that coarse segmentation obtains, for V component, adopt the OTSU Threshold Segmentation Algorithm that segmentation effect is stable, adaptivity is strong to carry out image subdivision and cut, thus reject the non-fiber pixel that part is examined by mistake.

Wherein, the mathematical description of OTSU algorithm is as follows: establish Probability p (i)=ni/N that brightness value i in image-region occurs (i=0,1 ..., 255), N is image-region interior pixel sum, the pixel number of ni to be brightness value be i. For threshold value t, image-region interior pixel being divided into target (M) and background (B) two class, target brightness value is greater than background luminance value.

Note: $w_B\left(t\right)=\underset{i=0}{\overset{t}{\Σ}}{n}_i/N;$

$w_M\left(t\right)=\underset{i=t+1}{\overset{255}{\Σ}}{n}_i/N;$

${\mathrm{\μ}}_B\left(t\right)=(\underset{i=0}{\overset{t}{\Σ}}i\·p(i\left)\right)/w_B\left(t\right);$

${\mathrm{\μ}}_M\left(t\right)=(\underset{i=t+1}{\overset{255}{\Σ}}i\·p(i\left)\right)/w_M\left(t\right);$

Then inter-class variance �� (t) can represent and is:

�� (t)=w_B(t)w_M(t)(��_B(t)-��_M(t))²;

Travel through all gray-scale values, choose the threshold value t making �� (t) maximum, that is:

$t=Arg\left\{\underset{100\≤x\≤255}{max}\mathrm{\σ}\left(t\right)\right\}$

Generally, in similar collection structure, fiber brightness is all more than 100, and in order to reduce segmentation and time loss by mistake, the lower limit arranging fiber brightness is 100, and now threshold value t value is:

$t=Arg\left\{\underset{100\≤x\≤255}{max}\mathrm{\σ}\left(t\right)\right\}$

According to threshold value t, zone of fiber being carried out segmentation and cuts, it is object point that brightness value is greater than the pixel of t, the non-fiber pixel that rejecting part is examined by mistake.

S303, carries out fiber target extraction to the figure after segmentation.

For the image after segmentation, the method labeled fibers target that in employing, value filtering and mathematics morphology combine, step is as follows:

S3031, in using for the image after segmentation, value filtering eliminates isolated noise point;

S3032, adopts the operation of the dilation and erosion in mathematics morphology, removes " hole " of image;

S3033, adopts the adjacent connection method of 8-, is identified by the binary map picture detected. In addition, the pixel count comprised due to fiber target (is generally 40��200, unit: pixel) in a limited scope, it is possible to adopting dual-threshold voltage to reject false target, bottom threshold is set to 40, and the upper limit is set to 200. Wherein, dilation and erosion operation and the adjacent connection method of 8-in middle value filtering, mathematics morphology are approach well known, have just repeated no more at this.

S304, obtains the fiber that same bar fiber presents respectively in FIMG and BIMG, is mated by the fiber in FIMG and BIMG, if two fibers mate mutually, bill to be detected is by double-colored fiber check and measure; Otherwise, bill to be detected is false tickets.

Same bar fiber can present red blue contrast effect in FIMG and BIMG. Therefore, before carrying out double-colored Characteristics Detection, first being mated by the fiber in FIMG and BIMG, which bar fiber is also exactly certain fiber found out in FIMG corresponding in BIMG is.

The maximum enclosure rectangle assuming certain fiber in image FIMG is:

$R_{FIMG}=\left\{(x,y)\right|x_{\min }^{FIMG}<x<x_{\max }^{FIMG},y_{\min }^{FIMG}<y<y_{\max }^{FIMG}\}$

Wherein, (x, y) is any pixel point on this fiber.

With reason, it is assumed that the maximum enclosure rectangle of certain fiber in image BIMG is:

$R_{BIMG}=\left\{(x,y)\right|x_{\min }^{BIMG}<x<x_{\max }^{BIMG},y_{\min }^{BIMG}<y<y_{\max }^{BIMG}\}$

Here two functions are defined:

Wherein, whether function Bound is for describing point (x, y) at rectangle R₁=(x, y) | x₁��x��x₂,y₁��y��y₂In; Function Cross is for describing rectangle R₂=(x, y) | x_l��x��x_r,y_t��y��y_bAnd rectangle R₁=(x, y) | x₁��x��x₂,y₁��y��y₂Whether in cross-shaped state.

Owing to same fiber meets horizontal conditions mirror in the position of FIMG and BIMG, therefore, judgement function it is constructed as follows here:

$\begin{array}{c}f=Bound(W-x_{\min }^{BIMG},y_{\min }^{BIMG},x_{\min }^{FIMG},x_{\max }^{FIMG},y_{\min }^{FIMG},y_{\max }^{FIMG})\\ +Bound(W-x_{\min }^{BIMG},y_{\max }^{BIMG},x_{\min }^{FIMG},x_{\max }^{FIMG},y_{\min }^{FIMG},y_{\max }^{FIMG})\\ +Bound(W-x_{\max }^{BIMG},y_{\min }^{BIMG},x_{\min }^{FIMG},x_{\max }^{FIMG},y_{\min }^{FIMG},y_{\max }^{FIMG})\\ +Bound(W-x_{\max }^{BIMG},y_{\max }^{BIMG},x_{\min }^{FIMG},x_{\max }^{FIMG},y_{\min }^{FIMG},y_{\max }^{FIMG})\\ +Bound(W-x_{\min }^{FIMG},y_{\min }^{FIMG},x_{\min }^{BIMG},x_{\max }^{BIMG},y_{\min }^{BIMG},y_{\max }^{BIMG})\\ +Bound(W-x_{\min }^{FIMG},y_{\max }^{FIMG},x_{\min }^{BIMG},x_{\max }^{BIMG},y_{\min }^{BIMG},y_{\max }^{BIMG})\\ +Bound(W-x_{\max }^{FIMG},y_{\min }^{FIMG},x_{\min }^{BIMG},x_{\max }^{BIMG},y_{\min }^{BIMG},y_{\max }^{BIMG})\\ +Bound(W-x_{\max }^{FIMG},y_{\max }^{FIMG},x_{\min }^{BIMG},x_{\max }^{BIMG},y_{\min }^{BIMG},y_{\max }^{BIMG})\\ +Cross(W-x_{\min }^{FIMG},W-x_{\max }^{FIMG},y_{\min }^{FIMG},y_{\max }^{FIMG},x_{\min }^{BIMG},x_{\max }^{BIMG},y_{\min }^{BIMG},y_{\max }^{BIMG})\\ +Cross(W-x_{\min }^{BIMG},W-x_{\max }^{BIMG},y_{\min }^{BIMG},y_{\max }^{BIMG},x_{\min }^{FIMG},x_{\max }^{FIMG},y_{\min }^{FIMG},y_{\max }^{FIMG})\end{array}$

Wherein, W represents the width of image.

If f is not zero, showing that two fibers mate mutually, bill to be detected is by double-colored fiber check and measure; Otherwise, show that two fibers do not mate.

In embodiment provided by the present invention, when carrying out the true and false of double-colored fiber check and measure bill to be detected, it is also possible to adjudicated the true and false of bill to be detected by double-colored characteristic. For same bar fiber, extract the minimum value of its H component in FIMG and BIMG, mean value and maximum value respectively, it is designated as respectively

The degree of depth embedding paper due to fiber is different, therefore on fiber, the H value of each pixel also has deviation. Therefore, when adjudicating double-colored characteristic, can not require that all pixels on fiber all meet double-colored characteristic, can not require that all fibres in bill all meets double-colored characteristic. Here the double-colored characteristic judgement function built is:

If F is not zero, showing that this fiber meets double-colored characteristic, bill to be detected is true; Otherwise, showing that this fiber does not meet double-colored characteristic, bill to be detected is false.

In the image of bill to be detected, detect all position of fibers and the shape that meet double-colored characteristic, it is possible to as the foundation of follow-up further Fibre sorting.

Gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, by phosphor pattern detection, the graphic information gathered is detected, specifically comprise the steps:

S311, control light source, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and converts thereof into the picture of HSV form.

S312, to the V component of the ultraviolet reflectance image FIMG in front, carries out little wave conversion, retains the graphic information of the low-frequency information FIMG0 after conversion as subsequent disposal.

The contribution of little wave conversion has two aspects, and one is the size reducing pending image under the prerequisite retaining the main information of image, thus reduces the computing amount that successive image processes each stage; Two is the interference reducing illumination and noise, strengthens the robustness of fingerprint characteristic detection method. In embodiment provided by the present invention, the formula of little wave conversion used is:

Wherein, W (j, m, n) coefficient is being similar at yardstick j place's image f (x, y), and �� is Haar wavelet scaling function, is formulated as:

S313, to the V component of low-frequency information FIMG0, adopts OTSU Threshold Segmentation Algorithm to carry out image subdivision and cuts.

To the V component of FIMG0, adopt aforesaid OTSU Threshold Segmentation Algorithm to carry out image subdivision and cut. Equally, in similar collection structure, the brightness of each pixel of phosphor pattern is all more than 100, and in order to reduce segmentation and time loss by mistake, arranging brightness lower limit is 100, and now threshold value t value is:

$t=Arg\left\{\underset{100\&le;x\&le;255}{max}\mathrm{\&sigma;}\left(t\right)\right\}$

When splitting, it is object point that brightness value is greater than the pixel of t, in the binary map picture after splitting according to threshold value t, not only comprises pattern, and comprises part fiber. Owing to fiber target is very little relative to pattern target, the impact of pattern characteristics detected result is very little, therefore without the need to rejecting fiber target.

S314, extracts the fingerprint characteristic of sola bill image of standard from cloud platform, and the phosphor pattern feature of image after segmentation being cut is mated with the phosphor pattern feature of the sola bill image of standard, true and false by the value decides bill of relation conefficient.

The fingerprint characteristic of the sola bill image of standard is extracted from cloud platform, the two-value template image assuming the phosphor pattern of the sola bill image of standard under identical size is MIMG, in embodiment provided by the present invention, adopting relevant matches method to carry out template matches, the calculation formula of relation conefficient is:

$r=\frac{\underset{y=1}{\overset{H}{\&Sigma;}}\underset{x=1}{\overset{W}{\&Sigma;}}FIMG0(x,y)\&CenterDot;MIMG(x,y)}{\sqrt{\underset{y=1}{\overset{H}{\&Sigma;}}\underset{x=1}{\overset{W}{\&Sigma;}}FIMG0{(x,y)}^2}\&CenterDot;\sqrt{\underset{y=1}{\overset{H}{\&Sigma;}}\underset{x=1}{\overset{W}{\&Sigma;}}MIMG{(x,y)}^2}}$

Owing to image to be matched (segmentation cut after image) and template image (the sola bill image of standard) are all that binary map picture, background and target represent with 0 and 1 respectively. Therefore r can carry out calculating fast with following formula:

Wherein, & represents AND operation.

Setting phosphor pattern detection threshold, in embodiment provided by the present invention, phosphor pattern detection threshold is set to 70, if r > 70, judges that phosphor pattern meets false proof requirement; Otherwise, judging that phosphor pattern does not meet false proof requirement, bill is false tickets.

Gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, by infrared image blanking feature detection, the graphic information gathered is detected, specifically comprise the steps:

S321, control light source, gathers the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected respectively, and it is passed through formula:Be converted to gray-scale map picture.

Control light source, gathers the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected respectively. The image gathered is RGB24 form, is converted into gray-scale map picture, and formula is as follows:

$V=\frac{1}{3}(R+G+B)$

Wherein, RGB represents the color of red, green, blue three passages respectively, and V represents the brightness in hsv color space (V).

S322, to the gray-scale map picture of the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected, carries out little wave conversion respectively, retains the graphic information of the low-frequency information RIMG0 after conversion and WIMG0 as subsequent disposal.

To the gray-scale map picture of the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected, carry out little wave conversion respectively, retain the graphic information of the low-frequency information RIMG0 after conversion and WIMG0 as subsequent disposal. The effect of little wave conversion is identical with description above with implementation method. Just repeat no more at this.

S323, builds corresponding threshold binary image TIMG respectively, carries out threshold binary image segmentation low-frequency information RIMG0 and WIMG0, obtains corresponding binary map as RIMG1 and WIMG1.

From intensity profile, the intensity profile of the targets such as line frame, printing type face, hand-written script is variant, remarkable not with the gray difference of the backgrounds such as bill paper, it is difficult to effectively divide the targets such as secant frame, font according to one or several threshold values. In embodiment provided by the present invention, it is proposed that a kind of new threshold binary image segmentation algorithm, it is possible to effectively divide the targets such as secant frame, font. Specifically comprise the steps:

S3231, builds corresponding threshold binary image TIMG respectively to low-frequency information RIMG0 and WIMG0. In embodiment provided by the present invention, for low-frequency information RIMG0 or WIMG0 (being abbreviated as f) here, build corresponding threshold binary image TIMG, step is as follows:

S32311, for any pixel point (x, y), calculates the gray average M1 in its neighborhood window;

S32312, judges inequality f (x, y) > whether M1 set up, if set up, proceeds to step S32313; Otherwise, proceed to step S32314;

S32313, threshold binary image TIMG (x, the y)=M1 at point (x, y) place;

S32314, calculates optimal segmenting threshold t, threshold binary image TIMG (x, the y)=t at point (x, y) place; Specifically comprise the steps:

First, calculate neighborhood window (11 �� 11) interior pixel intensity profile histogram p [i]=ni/N (i=0,1,, 255), the pixel number of ni to be brightness value be i, N is image-region interior pixel sum, N=11 �� 11=121 here;

Then, according to histogrammic first the valley point M2 (as shown in Figure 5) of direction search of gray-scale value from M1 to 2; M2 satisfies condition (p [M2+1] < p [M2-2] and p [M2-1] < p [M2+2]);

Judge whether M1's and M2 is equal, if M2=M1, then TIMG (x, y)=M1; Otherwise, between M1 and M2, utilize OTSU algorithms selection optimal segmenting threshold t:

$t=Arg\left\{\underset{M_2\&le;t\&le;M_1}{max}\mathrm{\&sigma;}\left(t\right)\right\}$

Wherein, address before utilizing OTSU algorithms selection optimal segmenting threshold t, just repeated no more at this.

Now TIMG (x, y)=t. S3232, adopts following formula to carry out threshold binary image segmentation, obtains corresponding binary map as RIMG1 and WIMG1.

Like this, grey image R IMG0 and WIMG0 is after over-segmentation, and corresponding binary map picture is designated as RIMG1 and WIMG1 respectively.

S324, merges binary map as RIMG1 and WIMG1, obtains binary map as RWIMG1:

Object point in RWIMG1 is the frame of blanking under infrared image and font mainly.

S325, the two-value template image MIMG2 of the infrared image of the sola bill image of standard is extracted from cloud platform, correlation matching algorithm is adopted binary map to be mated with two-value template image MIMG2 as RWIMG1, if the match is successful, then the infrared blanking characteristic of band detection bill meets false proof requirement, and bill is true; Otherwise, judging that the infrared blanking characteristic of band detection bill does not meet false proof requirement, bill is false.

Obtain the fingerprint characteristic of the sola bill image of standard from cloud platform, it is assumed that under identical size, the two-value template image of the infrared image of the sola bill image of standard is MIMG2, in embodiment provided by the present invention, adopt relevant matches method to carry out template matches,

The calculation formula of relation conefficient is:

Setting phosphor pattern detection threshold, in embodiment provided by the present invention, phosphor pattern detection threshold is set to 70, if r > 70, judges that infrared blanking characteristic meets false proof requirement, and bill is true; Otherwise, judging that infrared blanking characteristic does not meet false proof requirement, bill is false.

S4, when three kinds of detections all judge that bill to be detected is as, time true, judging that bill to be detected is as true, otherwise, judge that bill to be detected is as vacation.

In sum, the bank money fingerprint characteristic anti false authentication method based on cloud platform provided by the present invention, opens greatly the multispectral image of bill by camera acquisition; The multispectral image of the big bill gathered is carried out pre-treatment, obtain the sola bill image of standard, and the fingerprint characteristic of all sola bill images is stored into cloud platform, effectively ensure that the accuracy of the sola bill image of standard and comprehensive, improve the comprehensive of anti-counterfeiting detection Data Source, and then improve the accuracy of detection, it is stored into cloud platform and effectively saves local storage space. In addition, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, respectively the graphic information gathered is detected by double-colored fiber check and measure, phosphor pattern detection, infrared image blanking feature detection; When three kinds of detections all judge that bill to be detected is as, time true, judging that bill to be detected is as true, when three kinds of detections have one to judge bill to be detected as vacation, judge that bill to be detected is as vacation. This method eliminates the drawback that single bill authentication technique carries out false proof qualification, effectively improve the accuracy of false proof qualification.

Above the bank money fingerprint characteristic anti false authentication method based on cloud platform provided by the present invention is described in detail. For one of ordinary skill in the art, any apparent change it done under the prerequisite not deviating from true spirit, all by forming the infringement to patent right of the present invention, will undertake corresponding legal obligation.

Claims (10)

1. the bank money fingerprint characteristic anti false authentication method based on cloud platform, it is characterised in that comprise the steps:

S1, gathers the multispectral image of a big bill;

S2, carries out pre-treatment to described multispectral image, obtains the sola bill image of standard, and the fingerprint characteristic of all sola bill images is stored into cloud platform;

S3, gathers the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, described graphic information is detected respectively by double-colored fiber check and measure, phosphor pattern detection, infrared image blanking feature detection;

S4, when three kinds of detections all judge that bill to be detected is as, time true, judging that bill to be detected is as true, otherwise, judge that bill to be detected is as vacation.

2. as claimed in claim 1 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterised in that in step s 2, described multispectral image is carried out pre-treatment, obtain the sola bill image of standard, comprise the steps:

S21, according to distortion factor and distortion model, carries out distortion correction to the multispectral image of the big bill gathered;

S22, to the image carrying out obtaining after distortion corrects, carries out the sola bill image that scaling cutting obtains standard.

3., as claimed in claim 2 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterised in that in step S22, to the image carrying out obtaining after distortion corrects, carry out the sola bill image that scaling cutting obtains standard, comprise the steps:

S221, for the white light reflection image WIMG of a big bill, is transformed into hsv color space by RGB color;

S222, adopt fixed threshold dividing method, using H, S component in white light reflection image WIMG respectively the pixel in the first fixed interval as object point, obtain binary segmentation image;

S223, adopts the rectangular window identical with bill water solvus size to be scanned by image, the object point number in statistics rectangular window, if object point data exceed the 1/3 of rectangular window size, water solvus position then being detected, record rectangular window central position is (x0, y0);

S224, with coordinate point (x0-w1, y0-h1) it is bill starting point, the cutting of sola bill image is carried out according to bill size, obtain the sola bill image of standard, wherein, w1 is the distance on the left of bill water solvus regional center distance bill, and h1 is the distance on the upside of bill water solvus regional center distance bill.

4. as claimed in claim 1 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterized in that in step s3, gather the graphic information of bill to be detected, by double-colored fiber check and measure, the graphic information gathered is detected, comprise the steps:

S301, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and converts thereof into the picture of HSV form;

S302, according to the feature of blue fiber red in bill, using H, S component in the picture of HSV form respectively the pixel in the 2nd fixed interval as object point, carry out H, S component coarse segmentation, V component adopts OTSU Threshold Segmentation Algorithm carry out image subdivision and cuts;

S303, carries out fiber target extraction to the figure after segmentation;

S304, obtain the fiber that same bar fiber presents respectively in the ultraviolet reflectance image FIMG in front and the ultraviolet reflectance image BIMG at the back side, fiber in the ultraviolet reflectance image FIMG in front and the ultraviolet reflectance image BIMG at the back side is mated, if two bar fiber mates mutually, bill to be detected is by double-colored fiber check and measure; Otherwise, bill to be detected is false tickets.

5. as claimed in claim 1 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterised in that:

When carrying out the true and false of double-colored fiber check and measure bill to be detected, adjudicated the true and false of bill to be detected further by double-colored characteristic.

6. as claimed in claim 5 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterised in that the described true and false adjudicating bill to be detected by double-colored characteristic comprises the steps:

For same bar fiber, extract the minimum value of its H component in the ultraviolet reflectance image FIMG in front and the ultraviolet reflectance image BIMG at the back side, mean value and maximum value respectively, it is designated as respectively $(H_{\min }^{FIMG},H_{mean}^{FIMG},H_{max}^{FIMG}),(H_{\min }^{BIMG},H_{mean}^{BIMG},H_{\max }^{BIMG});$

The double-colored characteristic judgement function built is:

Judge the value of F; If F is not zero, then described fiber meets double-colored characteristic, and bill to be detected is true; Otherwise, described fiber does not meet double-colored characteristic, and bill to be detected is false.

7. as claimed in claim 1 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterized in that in step s3, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, by phosphor pattern detection, the graphic information gathered is detected, comprise the steps:

S311, gathers the ultraviolet reflectance image FIMG in the front of the bill to be detected and ultraviolet reflectance image BIMG at the back side respectively, and converts thereof into the picture of HSV form;

S312, to the V component of the ultraviolet reflectance image FIMG in front, carries out little wave conversion, obtains the low-frequency information FIMG0 after conversion;

S313, to the V component of described low-frequency information FIMG0, adopts OTSU Threshold Segmentation Algorithm to carry out image subdivision and cuts;

S314, extracts the fingerprint characteristic of sola bill image of standard from cloud platform, and the phosphor pattern feature of image after segmentation being cut is mated with the phosphor pattern feature of the sola bill image of standard, true and false by the value decides bill of relation conefficient.

8. as claimed in claim 7 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterised in that:

In step S314, the calculation formula of described relation conefficient is:

Wherein, the low-frequency information of the ultraviolet reflectance image FIMG that FIMG0 (x, y) is front on V component; The two-value template image of the sola bill image that MIMG (x, y) is standard; When relation conefficient is greater than phosphor pattern detection threshold, judge that phosphor pattern meets false proof requirement; Bill is true; Otherwise, judging that phosphor pattern does not meet false proof requirement, bill is false.

9. as claimed in claim 1 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterized in that in step s3, gather the graphic information of bill to be detected, in conjunction with the fingerprint characteristic of the sola bill image that cloud platform stores, by infrared image blanking feature detection, the graphic information gathered is detected, comprise the steps:

S321, gathers the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected respectively, and is converted to gray-scale map picture;

S322, to the gray-scale map picture of the infrared external reflection image RIMG and white light reflection image WIMG in the front of bill to be detected, carries out little wave conversion respectively, obtains the low-frequency information RIMG0 after conversion and WIMG0;

S323, builds corresponding threshold binary image respectively, carries out threshold binary image segmentation low-frequency information RIMG0 and WIMG0, obtains corresponding binary map as RIMG1 and WIMG1;

S324, merges binary map as RIMG1 and WIMG1, obtains binary map as RWIMG1:

S325, extract the two-value template image of the infrared image of the sola bill image of standard from cloud platform, adopt correlation matching algorithm to be mated with described two-value template image by described binary map picture, if the match is successful, then the infrared blanking characteristic of band detection bill meets false proof requirement, and bill is true; Otherwise, judging that the infrared blanking characteristic of band detection bill does not meet false proof requirement, bill is false.

10., as claimed in claim 1 based on the bank money fingerprint characteristic anti false authentication method of cloud platform, it is characterised in that in step S323, low-frequency information RIMG0 is built threshold binary image, comprises the steps:

For any pixel point (x, y), calculate the gray average M1 in its neighborhood window;

Judge whether inequality R IMG0 (x, y) > M1 sets up, if set up, threshold binary image TIMG (x, the y)=M1 at point (x, y) place;

Otherwise, calculate optimal segmenting threshold t, threshold binary image TIMG (x, the y)=t at point (x, y) place.